[{"year":"2024","intvolume":"         6","citation":{"ama":"Frantar E, Alistarh D-A. QMoE: Sub-1-bit compression of trillion parameter models. In: <i>Proceedings of Machine Learning and Systems</i>. Vol 6. ; 2024.","ista":"Frantar E, Alistarh D-A. 2024. QMoE: Sub-1-bit compression of trillion parameter models. Proceedings of Machine Learning and Systems. MLSys: Machine Learning and Systems vol. 6.","ieee":"E. Frantar and D.-A. Alistarh, “QMoE: Sub-1-bit compression of trillion parameter models,” in <i>Proceedings of Machine Learning and Systems</i>, Santa Clara, CA, United States, 2024, vol. 6.","chicago":"Frantar, Elias, and Dan-Adrian Alistarh. “QMoE: Sub-1-Bit Compression of Trillion Parameter Models.” In <i>Proceedings of Machine Learning and Systems</i>, Vol. 6, 2024.","short":"E. Frantar, D.-A. Alistarh, in:, Proceedings of Machine Learning and Systems, 2024.","mla":"Frantar, Elias, and Dan-Adrian Alistarh. “QMoE: Sub-1-Bit Compression of Trillion Parameter Models.” <i>Proceedings of Machine Learning and Systems</i>, vol. 6, 2024.","apa":"Frantar, E., &#38; Alistarh, D.-A. (2024). QMoE: Sub-1-bit compression of trillion parameter models. In <i>Proceedings of Machine Learning and Systems</i> (Vol. 6). Santa Clara, CA, United States."},"day":"01","date_published":"2024-05-01T00:00:00Z","main_file_link":[{"open_access":"1","url":"https://proceedings.mlsys.org/paper_files/paper/2024/hash/c74b624843218d9b6713fcf299d6d5e4-Abstract-Conference.html"}],"article_processing_charge":"No","publication_status":"published","related_material":{"record":[{"id":"17485","relation":"dissertation_contains","status":"public"}]},"department":[{"_id":"DaAl"}],"author":[{"full_name":"Frantar, Elias","first_name":"Elias","last_name":"Frantar","id":"09a8f98d-ec99-11ea-ae11-c063a7b7fe5f"},{"orcid":"0000-0003-3650-940X","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","first_name":"Dan-Adrian","last_name":"Alistarh","full_name":"Alistarh, Dan-Adrian"}],"oa_version":"Published Version","conference":{"name":"MLSys: Machine Learning and Systems","end_date":"2024-05-16","location":"Santa Clara, CA, United States","start_date":"2024-05-13"},"publication":"Proceedings of Machine Learning and Systems","ddc":["000"],"language":[{"iso":"eng"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","date_updated":"2026-07-29T13:48:40Z","status":"public","_id":"18061","title":"QMoE: Sub-1-bit compression of trillion parameter models","quality_controlled":"1","abstract":[{"text":"Mixture-of-Experts (MoE) architectures offer a general solution to the high inference costs of large language models (LLMs) via sparse routing, bringing faster and more accurate models, at the cost of massive parameter counts. For example, the SwitchTransformer-c2048 model has 1.6 trillion parameters, requiring 3.2TB of accelerator memory to run efficiently, which makes practical deployment challenging and expensive. In this paper, we present a solution to this memory problem, in form of a new compression and execution framework called QMoE. Specifically, QMoE consists of a scalable algorithm which accurately compresses trillion-parameter MoEs to less than 1 bit per parameter, in a custom format co-designed with bespoke GPU decoding kernels to facilitate efficient end-to-end compressed inference, with minor runtime overheads relative to uncompressed execution. Concretely, QMoE can compress the 1.6 trillion parameter SwitchTransformer-c2048 model to less than 160GB (20x compression, 0.8 bits per parameter) at only minor accuracy loss, in less than a day on a single GPU. This enables, for the first time, the execution of a trillion-parameter model on affordable commodity hardware, like a single server with 4x NVIDIA A6000 or 8x NVIDIA 3090 GPUs, at less than 5% runtime overhead relative to ideal uncompressed inference. The anonymized code is available at: github.com/mlsys24-qmoe/qmoe.","lang":"eng"}],"date_created":"2024-09-13T10:01:38Z","oa":1,"month":"05","type":"conference","das_tickbox":"1","corr_author":"1","volume":6},{"publication_status":"published","page":"232","project":[{"name":"Prix Lopez-Loretta 2019 - Marco Mondelli","_id":"059876FA-7A3F-11EA-A408-12923DDC885E"},{"_id":"9B9290DE-BA93-11EA-9121-9846C619BF3A","grant_number":"W1260-N35","name":"Vienna Graduate School on Computational Optimization"}],"year":"2024","date_updated":"2026-06-18T17:55:53Z","oa_version":"Published Version","date_created":"2024-08-28T15:14:25Z","oa":1,"abstract":[{"lang":"eng","text":"In the modern age of machine learning, artificial neural networks have become an integral part\r\nof many practical systems. One of the key ingredients of the success of the deep learning\r\napproach is recent computational advances which allowed the training of models with billions\r\nof parameters on large-scale data. Such over-parameterized and data-hungry regimes pose a\r\nchallenge for the theoretical analysis of modern models since “classical” statistical wisdom\r\nis no longer applicable. In this view, it is paramount to extend or develop new machinery\r\nthat will allow tackling the neural network analysis under new challenging asymptotic regimes,\r\nwhich is the focus of this thesis.\r\nLarge neural network systems are usually optimized via “local” search algorithms, such\r\nas stochastic gradient descent (SGD). However, given the high-dimensional nature of the\r\nparameter space, it is a priori not clear why such a crude “local” approach works so remarkably\r\nwell in practice. We take a step towards demystifying this phenomenon by showing that\r\nthe landscape of the SGD training dynamics exhibits a few beneficial properties for the\r\noptimization. First, we show that along the SGD trajectory an over-parameterized network\r\nis dropout stable. The emergence of dropout stability allows to conclude that the minima\r\nfound by SGD are connected via a continuous path of small loss. This in turn means that\r\nthe high-dimensional landscape of the neural network optimization problem is provably not so\r\nunfavourable to gradient-based training, due to mode connectivity. Next, we show that SGD\r\nfor an over-parameterized network tends to find solutions that are functionally more “simple”.\r\nThis in turn means that the SGD minima are more robust, since a less complicated solution\r\nwill less likely overfit the data. More formally, for a prototypical example of a wide two-layer\r\nReLU network on a 1d regression task we show that the SGD algorithm is implicitly selective in\r\nits choice of an interpolating solution. Namely, at convergence the neural network implements\r\na piece-wise linear function with the number of linear regions depending only on the amount\r\nof training data. This is in contrast to a “smooth”-like behaviour which one would expect\r\ngiven such a severe over-parameterization of the model.\r\nDiverging from the generic supervised setting of classification and regression problems, we\r\nanalyze an auto-encoder model that is commonly used for representation learning and data\r\ncompression. Despite the wide applicability of the auto-encoding paradigm, the theoretical\r\nunderstanding of their behaviour is limited even in the simplistic shallow case. The related\r\nwork is restricted to extreme asymptotic regimes in which the auto-encoder is either severely\r\nover-parameterized or under-parameterized. In contrast, we provide a tight characterization\r\nfor the 1-bit compression of Gaussian signals in the challenging proportional regime, i.e., the\r\ninput dimension and the size of the compressed representation obey the same asymptotics.\r\nWe also show that gradient-based methods are able to find a globally optimal solution and\r\nthat the predictions made for Gaussian data extrapolate beyond - to the case of compression\r\nof natural images. Next, we relax the Gaussian assumption and study more structured input\r\nsources. We show that the shallow model is sometimes agnostic to the structure of the data\r\nvii\r\nwhich results in a Gaussian-like behaviour. We prove that making the decoding component\r\nslightly less shallow is already enough to escape the “curse” of Gaussian performance.\r\n"}],"title":"High-dimensional limits in artificial neural networks","_id":"17465","status":"public","supervisor":[{"first_name":"Marco","last_name":"Mondelli","full_name":"Mondelli, Marco","orcid":"0000-0002-3242-7020","id":"27EB676C-8706-11E9-9510-7717E6697425"},{"full_name":"Alistarh, Dan-Adrian","first_name":"Dan-Adrian","last_name":"Alistarh","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-3650-940X"}],"corr_author":"1","degree_awarded":"PhD","file_date_updated":"2024-10-05T22:30:05Z","article_processing_charge":"No","date_published":"2024-08-29T00:00:00Z","day":"29","citation":{"short":"A. Shevchenko, High-Dimensional Limits in Artificial Neural Networks, Institute of Science and Technology Austria, 2024.","ama":"Shevchenko A. High-dimensional limits in artificial neural networks. 2024. doi:<a href=\"https://doi.org/10.15479/at:ista:17465\">10.15479/at:ista:17465</a>","ieee":"A. Shevchenko, “High-dimensional limits in artificial neural networks,” Institute of Science and Technology Austria, 2024.","ista":"Shevchenko A. 2024. High-dimensional limits in artificial neural networks. Institute of Science and Technology Austria.","chicago":"Shevchenko, Alexander. “High-Dimensional Limits in Artificial Neural Networks.” Institute of Science and Technology Austria, 2024. <a href=\"https://doi.org/10.15479/at:ista:17465\">https://doi.org/10.15479/at:ista:17465</a>.","mla":"Shevchenko, Alexander. <i>High-Dimensional Limits in Artificial Neural Networks</i>. Institute of Science and Technology Austria, 2024, doi:<a href=\"https://doi.org/10.15479/at:ista:17465\">10.15479/at:ista:17465</a>.","apa":"Shevchenko, A. (2024). <i>High-dimensional limits in artificial neural networks</i>. Institute of Science and Technology Austria. <a href=\"https://doi.org/10.15479/at:ista:17465\">https://doi.org/10.15479/at:ista:17465</a>"},"publisher":"Institute of Science and Technology Austria","file":[{"relation":"main_file","checksum":"da6dd3166078934577f6af93d27000e2","file_size":4468610,"access_level":"open_access","date_updated":"2024-10-05T22:30:05Z","content_type":"application/pdf","creator":"ashevche","file_name":"thesis_a2b.pdf","file_id":"17482","embargo":"2024-10-04","date_created":"2024-09-02T09:23:32Z"},{"file_id":"17483","file_name":"Thesis Alex - ISTA.zip","creator":"ashevche","date_created":"2024-09-02T09:23:46Z","embargo_to":"open_access","relation":"source_file","checksum":"76a39ef252239560923cdda4ce0a31a4","file_size":15930999,"access_level":"closed","date_updated":"2024-10-05T22:30:05Z","content_type":"application/zip"}],"user_id":"8b945eb4-e2f2-11eb-945a-df72226e66a9","ddc":["519"],"language":[{"iso":"eng"}],"acknowledged_ssus":[{"_id":"ScienComp"}],"OA_place":"repository","author":[{"id":"F2B06EC2-C99E-11E9-89F0-752EE6697425","first_name":"Aleksandr","last_name":"Shevchenko","full_name":"Shevchenko, Aleksandr"}],"department":[{"_id":"GradSch"},{"_id":"DaAl"},{"_id":"MaMo"}],"related_material":{"record":[{"id":"11420","relation":"part_of_dissertation","status":"public"},{"id":"14459","status":"public","relation":"part_of_dissertation"},{"relation":"part_of_dissertation","status":"public","id":"9198"},{"relation":"part_of_dissertation","status":"public","id":"17469"}]},"month":"08","publication_identifier":{"issn":["2663-337X"]},"alternative_title":["ISTA Thesis"],"has_accepted_license":"1","doi":"10.15479/at:ista:17465","type":"dissertation"},{"alternative_title":["PMLR"],"month":"07","quality_controlled":"1","external_id":{"arxiv":["2402.05013"]},"type":"conference","volume":235,"citation":{"apa":"Kögler, K., Shevchenko, A., Hassani, H., &#38; Mondelli, M. (2024). Compression of structured data with autoencoders: Provable benefit of nonlinearities and depth. In <i>Proceedings of the 41st International Conference on Machine Learning</i> (Vol. 235, pp. 24964–25015). Vienna, Austria: ML Research Press.","mla":"Kögler, Kevin, et al. “Compression of Structured Data with Autoencoders: Provable Benefit of Nonlinearities and Depth.” <i>Proceedings of the 41st International Conference on Machine Learning</i>, vol. 235, ML Research Press, 2024, pp. 24964–5015.","ista":"Kögler K, Shevchenko A, Hassani H, Mondelli M. 2024. Compression of structured data with autoencoders: Provable benefit of nonlinearities and depth. Proceedings of the 41st International Conference on Machine Learning. ICML: International Conference on Machine Learning, PMLR, vol. 235, 24964–25015.","chicago":"Kögler, Kevin, Alexander Shevchenko, Hamed Hassani, and Marco Mondelli. “Compression of Structured Data with Autoencoders: Provable Benefit of Nonlinearities and Depth.” In <i>Proceedings of the 41st International Conference on Machine Learning</i>, 235:24964–15. ML Research Press, 2024.","ieee":"K. Kögler, A. Shevchenko, H. Hassani, and M. Mondelli, “Compression of structured data with autoencoders: Provable benefit of nonlinearities and depth,” in <i>Proceedings of the 41st International Conference on Machine Learning</i>, Vienna, Austria, 2024, vol. 235, pp. 24964–25015.","ama":"Kögler K, Shevchenko A, Hassani H, Mondelli M. Compression of structured data with autoencoders: Provable benefit of nonlinearities and depth. In: <i>Proceedings of the 41st International Conference on Machine Learning</i>. Vol 235. ML Research Press; 2024:24964-25015.","short":"K. Kögler, A. Shevchenko, H. Hassani, M. Mondelli, in:, Proceedings of the 41st International Conference on Machine Learning, ML Research Press, 2024, pp. 24964–25015."},"publisher":"ML Research Press","intvolume":"       235","article_processing_charge":"No","date_published":"2024-07-01T00:00:00Z","main_file_link":[{"url":"https://proceedings.mlr.press/v235/kogler24a.html","open_access":"1"}],"day":"01","author":[{"id":"94ec913c-dc85-11ea-9058-e5051ab2428b","full_name":"Kögler, Kevin","first_name":"Kevin","last_name":"Kögler"},{"id":"F2B06EC2-C99E-11E9-89F0-752EE6697425","first_name":"Aleksandr","last_name":"Shevchenko","full_name":"Shevchenko, Aleksandr"},{"full_name":"Hassani, Hamed","first_name":"Hamed","last_name":"Hassani"},{"full_name":"Mondelli, Marco","last_name":"Mondelli","first_name":"Marco","id":"27EB676C-8706-11E9-9510-7717E6697425","orcid":"0000-0002-3242-7020"}],"department":[{"_id":"DaAl"},{"_id":"MaMo"}],"related_material":{"record":[{"status":"public","relation":"dissertation_contains","id":"17465"}]},"acknowledgement":"Kevin Kogler, Alexander Shevchenko and Marco Mondelli are supported by the 2019 Lopez-Loreta Prize. Hamed\r\nHassani acknowledges the support by the NSF CIF award (1910056) and the NSF Institute for CORE Emerging Methods in Data Science (EnCORE).","arxiv":1,"ddc":["000"],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","language":[{"iso":"eng"}],"publication":"Proceedings of the 41st International Conference on Machine Learning","title":"Compression of structured data with autoencoders: Provable benefit of nonlinearities and depth","status":"public","_id":"17469","date_created":"2024-08-29T11:47:57Z","oa":1,"abstract":[{"lang":"eng","text":"Autoencoders are a prominent model in many empirical branches of machine learning and lossy data compression. However, basic theoretical questions remain unanswered even in a shallow two-layer setting. In particular, to what degree does a shallow autoencoder capture the structure of the underlying data distribution? For the prototypical case of the 1-bit compression of sparse Gaussian data, we prove that gradient descent converges to a solution that completely disregards the sparse structure of the input. Namely, the performance of the algorithm is the same as if it was compressing a Gaussian source - with no sparsity. For general data distributions, we give evidence of a phase transition phenomenon in the shape of the gradient descent minimizer, as a function of the data sparsity: below the critical sparsity level, the minimizer is a rotation taken uniformly at random (just like in the compression of non-sparse data); above the critical sparsity, the minimizer is the identity (up to a permutation). Finally, by exploiting a connection with approximate message passing algorithms, we show how to improve upon Gaussian performance for the compression of sparse data: adding a denoising function to a shallow architecture already reduces the loss provably, and a suitable multi-layer decoder leads to a further improvement. We validate our findings on image datasets, such as CIFAR-10 and MNIST."}],"corr_author":"1","project":[{"name":"Prix Lopez-Loretta 2019 - Marco Mondelli","_id":"059876FA-7A3F-11EA-A408-12923DDC885E"}],"year":"2024","page":"24964-25015","publication_status":"published","scopus_import":"1","oa_version":"Published Version","date_updated":"2026-08-20T22:30:15Z","conference":{"location":"Vienna, Austria","start_date":"2024-07-21","end_date":"2024-07-27","name":"ICML: International Conference on Machine Learning"}},{"arxiv":1,"language":[{"iso":"eng"}],"publication":"Distributed Computing","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","author":[{"full_name":"Aksenov, Vitalii","last_name":"Aksenov","first_name":"Vitalii","id":"2980135A-F248-11E8-B48F-1D18A9856A87"},{"id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-3650-940X","full_name":"Alistarh, Dan-Adrian","first_name":"Dan-Adrian","last_name":"Alistarh"},{"last_name":"Drozdova","first_name":"Alexandra","full_name":"Drozdova, Alexandra"},{"full_name":"Mohtashami, Amirkeivan","last_name":"Mohtashami","first_name":"Amirkeivan"}],"department":[{"_id":"DaAl"}],"article_processing_charge":"No","main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2008.01009"}],"day":"01","date_published":"2023-09-01T00:00:00Z","citation":{"apa":"Aksenov, V., Alistarh, D.-A., Drozdova, A., &#38; Mohtashami, A. (2023). The splay-list: A distribution-adaptive concurrent skip-list. <i>Distributed Computing</i>. Springer Nature. <a href=\"https://doi.org/10.1007/s00446-022-00441-x\">https://doi.org/10.1007/s00446-022-00441-x</a>","mla":"Aksenov, Vitalii, et al. “The Splay-List: A Distribution-Adaptive Concurrent Skip-List.” <i>Distributed Computing</i>, vol. 36, Springer Nature, 2023, pp. 395–418, doi:<a href=\"https://doi.org/10.1007/s00446-022-00441-x\">10.1007/s00446-022-00441-x</a>.","ieee":"V. Aksenov, D.-A. Alistarh, A. Drozdova, and A. Mohtashami, “The splay-list: A distribution-adaptive concurrent skip-list,” <i>Distributed Computing</i>, vol. 36. Springer Nature, pp. 395–418, 2023.","chicago":"Aksenov, Vitalii, Dan-Adrian Alistarh, Alexandra Drozdova, and Amirkeivan Mohtashami. “The Splay-List: A Distribution-Adaptive Concurrent Skip-List.” <i>Distributed Computing</i>. Springer Nature, 2023. <a href=\"https://doi.org/10.1007/s00446-022-00441-x\">https://doi.org/10.1007/s00446-022-00441-x</a>.","ista":"Aksenov V, Alistarh D-A, Drozdova A, Mohtashami A. 2023. The splay-list: A distribution-adaptive concurrent skip-list. Distributed Computing. 36, 395–418.","ama":"Aksenov V, Alistarh D-A, Drozdova A, Mohtashami A. The splay-list: A distribution-adaptive concurrent skip-list. <i>Distributed Computing</i>. 2023;36:395-418. doi:<a href=\"https://doi.org/10.1007/s00446-022-00441-x\">10.1007/s00446-022-00441-x</a>","short":"V. Aksenov, D.-A. Alistarh, A. Drozdova, A. Mohtashami, Distributed Computing 36 (2023) 395–418."},"publisher":"Springer Nature","intvolume":"        36","volume":36,"doi":"10.1007/s00446-022-00441-x","external_id":{"isi":["000913424000001"],"arxiv":["2008.01009"]},"type":"journal_article","month":"09","quality_controlled":"1","article_type":"original","publication_identifier":{"eissn":["1432-0452"],"issn":["0178-2770"]},"date_updated":"2023-08-14T12:54:32Z","scopus_import":"1","oa_version":"Preprint","publication_status":"published","page":"395-418","year":"2023","isi":1,"date_created":"2023-01-22T23:00:55Z","oa":1,"abstract":[{"text":"The design and implementation of efficient concurrent data structures has seen significant attention. However, most of this work has focused on concurrent data structures providing good worst-case guarantees, although, in real workloads, objects are often accessed at different rates. Efficient distribution-adaptive data structures, such as splay-trees, are known in the sequential case; however, they often are hard to translate efficiently to the concurrent case. We investigate distribution-adaptive concurrent data structures, and propose a new design called the splay-list. At a high level, the splay-list is similar to a standard skip-list, with the key distinction that the height of each element adapts dynamically to its access rate: popular elements “move up,” whereas rarely-accessed elements decrease in height. We show that the splay-list provides order-optimal amortized complexity bounds for a subset of operations, while being amenable to efficient concurrent implementation. Experiments show that the splay-list can leverage distribution-adaptivity for performance, and can outperform the only previously-known distribution-adaptive concurrent design in certain workloads.","lang":"eng"}],"title":"The splay-list: A distribution-adaptive concurrent skip-list","status":"public","_id":"12330"},{"date_updated":"2025-04-14T07:49:16Z","scopus_import":"1","oa_version":"Published Version","license":"https://creativecommons.org/licenses/by/4.0/","publication_status":"published","year":"2023","project":[{"grant_number":"805223","name":"Elastic Coordination for Scalable Machine Learning","_id":"268A44D6-B435-11E9-9278-68D0E5697425","call_identifier":"H2020"},{"call_identifier":"H2020","_id":"26A5D39A-B435-11E9-9278-68D0E5697425","grant_number":"840605","name":"Coordination in constrained and natural distributed systems"}],"corr_author":"1","isi":1,"abstract":[{"text":"Approximate agreement is one of the few variants of consensus that can be solved in a wait-free manner in asynchronous systems where processes communicate by reading and writing to shared memory. In this work, we consider a natural generalisation of approximate agreement on arbitrary undirected connected graphs. Each process is given a node of the graph as input and, if non-faulty, must output a node such that\r\n– all the outputs are within distance 1 of one another, and\r\n– each output value lies on a shortest path between two input values.\r\nFrom prior work, it is known that there is no wait-free algorithm among  processes for this problem on any cycle of length , by reduction from 2-set agreement (Castañeda et al., 2018).\r\n\r\nIn this work, we investigate the solvability of this task on general graphs. We give a new, direct proof of the impossibility of approximate agreement on cycles of length , via a generalisation of Sperner's Lemma to convex polygons. We also extend the reduction from 2-set agreement to a larger class of graphs, showing that approximate agreement on these graphs is unsolvable. On the positive side, we present a wait-free algorithm for a different class of graphs, which properly contains the class of chordal graphs.","lang":"eng"}],"oa":1,"date_created":"2023-02-19T23:00:55Z","status":"public","_id":"12566","title":"Wait-free approximate agreement on graphs","tmp":{"name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","image":"/images/cc_by.png","short":"CC BY (4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"publication":"Theoretical Computer Science","ddc":["000"],"language":[{"iso":"eng"}],"user_id":"4359f0d1-fa6c-11eb-b949-802e58b17ae8","acknowledgement":"This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 805223 ScaleML) and under the Marie Skłodowska-Curie grant agreement No. 840605 and from the Natural Sciences and Engineering Research Council of Canada grant RGPIN-2020-04178. Part of this work was done while Faith Ellen was visiting IST Austria.","department":[{"_id":"DaAl"}],"ec_funded":1,"author":[{"id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-3650-940X","full_name":"Alistarh, Dan-Adrian","first_name":"Dan-Adrian","last_name":"Alistarh"},{"last_name":"Ellen","first_name":"Faith","full_name":"Ellen, Faith"},{"id":"334EFD2E-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0002-6432-6646","full_name":"Rybicki, Joel","first_name":"Joel","last_name":"Rybicki"}],"day":"28","date_published":"2023-02-28T00:00:00Z","article_processing_charge":"Yes (via OA deal)","file_date_updated":"2023-02-20T07:30:20Z","article_number":"113733","issue":"2","intvolume":"       948","publisher":"Elsevier","file":[{"file_id":"12570","file_name":"2023_TheoreticalCompScience_Alistarh.pdf","creator":"dernst","date_created":"2023-02-20T07:30:20Z","relation":"main_file","checksum":"b27c5290f2f1500c403494364ee39c9f","content_type":"application/pdf","success":1,"date_updated":"2023-02-20T07:30:20Z","access_level":"open_access","file_size":602333}],"citation":{"ama":"Alistarh D-A, Ellen F, Rybicki J. Wait-free approximate agreement on graphs. <i>Theoretical Computer Science</i>. 2023;948(2). doi:<a href=\"https://doi.org/10.1016/j.tcs.2023.113733\">10.1016/j.tcs.2023.113733</a>","ista":"Alistarh D-A, Ellen F, Rybicki J. 2023. Wait-free approximate agreement on graphs. Theoretical Computer Science. 948(2), 113733.","chicago":"Alistarh, Dan-Adrian, Faith Ellen, and Joel Rybicki. “Wait-Free Approximate Agreement on Graphs.” <i>Theoretical Computer Science</i>. Elsevier, 2023. <a href=\"https://doi.org/10.1016/j.tcs.2023.113733\">https://doi.org/10.1016/j.tcs.2023.113733</a>.","ieee":"D.-A. Alistarh, F. Ellen, and J. Rybicki, “Wait-free approximate agreement on graphs,” <i>Theoretical Computer Science</i>, vol. 948, no. 2. Elsevier, 2023.","short":"D.-A. Alistarh, F. Ellen, J. Rybicki, Theoretical Computer Science 948 (2023).","mla":"Alistarh, Dan-Adrian, et al. “Wait-Free Approximate Agreement on Graphs.” <i>Theoretical Computer Science</i>, vol. 948, no. 2, 113733, Elsevier, 2023, doi:<a href=\"https://doi.org/10.1016/j.tcs.2023.113733\">10.1016/j.tcs.2023.113733</a>.","apa":"Alistarh, D.-A., Ellen, F., &#38; Rybicki, J. (2023). Wait-free approximate agreement on graphs. <i>Theoretical Computer Science</i>. Elsevier. <a href=\"https://doi.org/10.1016/j.tcs.2023.113733\">https://doi.org/10.1016/j.tcs.2023.113733</a>"},"doi":"10.1016/j.tcs.2023.113733","has_accepted_license":"1","volume":948,"type":"journal_article","external_id":{"isi":["000934262700001"]},"article_type":"original","quality_controlled":"1","month":"02","publication_identifier":{"issn":["0304-3975"]}},{"publication_identifier":{"isbn":["9798400700156"]},"quality_controlled":"1","month":"02","type":"conference","external_id":{"arxiv":["2211.04986"]},"doi":"10.1145/3572848.3577481","publisher":"Association for Computing Machinery","citation":{"ieee":"N. Koval, D.-A. Alistarh, and R. Elizarov, “Fast and scalable channels in Kotlin Coroutines,” in <i>Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming</i>, Montreal, QC, Canada, 2023, pp. 107–118.","chicago":"Koval, Nikita, Dan-Adrian Alistarh, and Roman Elizarov. “Fast and Scalable Channels in Kotlin Coroutines.” In <i>Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming</i>, 107–18. Association for Computing Machinery, 2023. <a href=\"https://doi.org/10.1145/3572848.3577481\">https://doi.org/10.1145/3572848.3577481</a>.","ista":"Koval N, Alistarh D-A, Elizarov R. 2023. Fast and scalable channels in Kotlin Coroutines. Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming. PPoPP: Sympopsium on Principles and Practice of Parallel Programming, 107–118.","ama":"Koval N, Alistarh D-A, Elizarov R. Fast and scalable channels in Kotlin Coroutines. In: <i>Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming</i>. Association for Computing Machinery; 2023:107-118. doi:<a href=\"https://doi.org/10.1145/3572848.3577481\">10.1145/3572848.3577481</a>","short":"N. Koval, D.-A. Alistarh, R. Elizarov, in:, Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, Association for Computing Machinery, 2023, pp. 107–118.","apa":"Koval, N., Alistarh, D.-A., &#38; Elizarov, R. (2023). Fast and scalable channels in Kotlin Coroutines. In <i>Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming</i> (pp. 107–118). Montreal, QC, Canada: Association for Computing Machinery. <a href=\"https://doi.org/10.1145/3572848.3577481\">https://doi.org/10.1145/3572848.3577481</a>","mla":"Koval, Nikita, et al. “Fast and Scalable Channels in Kotlin Coroutines.” <i>Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming</i>, Association for Computing Machinery, 2023, pp. 107–18, doi:<a href=\"https://doi.org/10.1145/3572848.3577481\">10.1145/3572848.3577481</a>."},"main_file_link":[{"url":"https://doi.org/10.48550/arXiv.2211.04986","open_access":"1"}],"day":"25","date_published":"2023-02-25T00:00:00Z","article_processing_charge":"No","department":[{"_id":"DaAl"}],"author":[{"id":"2F4DB10C-F248-11E8-B48F-1D18A9856A87","first_name":"Nikita","last_name":"Koval","full_name":"Koval, Nikita"},{"full_name":"Alistarh, Dan-Adrian","first_name":"Dan-Adrian","last_name":"Alistarh","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-3650-940X"},{"full_name":"Elizarov, Roman","first_name":"Roman","last_name":"Elizarov"}],"publication":"Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","language":[{"iso":"eng"}],"arxiv":1,"_id":"12735","status":"public","title":"Fast and scalable channels in Kotlin Coroutines","abstract":[{"text":"Asynchronous programming has gained significant popularity over the last decade: support for this programming pattern is available in many popular languages via libraries and native language implementations, typically in the form of coroutines or the async/await construct. Instead of programming via shared memory, this concept assumes implicit synchronization through message passing. The key data structure enabling such communication is the rendezvous channel. Roughly, a rendezvous channel is a blocking queue of size zero, so both send(e) and receive() operations wait for each other, performing a rendezvous when they meet. To optimize the message passing pattern, channels are usually equipped with a fixed-size buffer, so sends do not suspend and put elements into the buffer until its capacity is exceeded. This primitive is known as a buffered channel.\r\n\r\nThis paper presents a fast and scalable algorithm for both rendezvous and buffered channels. Similarly to modern queues, our solution is based on an infinite array with two positional counters for send(e) and receive() operations, leveraging the unconditional Fetch-And-Add instruction to update them. Yet, the algorithm requires non-trivial modifications of this classic pattern, in order to support the full channel semantics, such as buffering and cancellation of waiting requests. We compare the performance of our solution to that of the Kotlin implementation, as well as against other academic proposals, showing up to 9.8× speedup. To showcase its expressiveness and performance, we also integrated the proposed algorithm into the standard Kotlin Coroutines library, replacing the previous channel implementations.","lang":"eng"}],"oa":1,"date_created":"2023-03-19T23:00:58Z","year":"2023","page":"107-118","publication_status":"published","oa_version":"Preprint","scopus_import":"1","conference":{"location":"Montreal, QC, Canada","start_date":"2023-02-25","end_date":"2023-03-01","name":"PPoPP: Sympopsium on Principles and Practice of Parallel Programming"},"date_updated":"2023-03-20T07:29:28Z"},{"oa":1,"date_created":"2023-03-19T23:00:58Z","abstract":[{"text":"Although a wide variety of handcrafted concurrent data structures have been proposed, there is considerable interest in universal approaches (Universal Constructions or UCs) for building concurrent data structures. UCs (semi-)automatically convert a sequential data structure into a concurrent one. The simplest approach uses locks [3, 6] that protect a sequential data structure and allow only one process to access it at a time. However, the resulting data structure is blocking. Most work on UCs instead focuses on obtaining non-blocking progress guarantees such as obstruction-freedom, lock-freedom or wait-freedom. Many non-blocking UCs have appeared. Key examples include the seminal wait-free UC [2] by Herlihy, a NUMA-aware UC [10] by Yi et al., and an efficient UC for large objects [1] by Fatourou et al.","lang":"eng"}],"status":"public","_id":"12736","title":"Unexpected scaling in path copying trees","conference":{"end_date":"2023-03-01","name":"PPoPP: Sympopsium on Principles and Practice of Parallel Programming","location":"Montreal, QB, Canada","start_date":"2023-02-25"},"date_updated":"2026-06-18T17:29:03Z","scopus_import":"1","oa_version":"Published Version","page":"438-440","publication_status":"published","year":"2023","doi":"10.1145/3572848.3577512","type":"conference_poster","month":"02","quality_controlled":"1","publication_identifier":{"isbn":["9798400700156"]},"language":[{"iso":"eng"}],"publication":"Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming","ddc":["000"],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","department":[{"_id":"DaAl"},{"_id":"GradSch"}],"acknowledgement":"This work was supported by: the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Program grant: RGPIN-2019-04227, and the Canada Foundation for Innovation John R. Evans Leaders Fund (CFI-JELF) with equal support from the Ontario Research Fund CFI Leaders Opportunity Fund: 38512.","author":[{"first_name":"Vitaly","last_name":"Aksenov","full_name":"Aksenov, Vitaly"},{"id":"3569F0A0-F248-11E8-B48F-1D18A9856A87","full_name":"Brown, Trevor A","last_name":"Brown","first_name":"Trevor A"},{"full_name":"Fedorov, Alexander","last_name":"Fedorov","first_name":"Alexander","id":"2e711909-896a-11ed-bdf8-eb0f5a2984c6"},{"last_name":"Kokorin","first_name":"Ilya","full_name":"Kokorin, Ilya"}],"article_processing_charge":"No","main_file_link":[{"url":"https://doi.org/10.1145/3572848.3577512","open_access":"1"}],"day":"25","date_published":"2023-02-25T00:00:00Z","citation":{"short":"V. Aksenov, T.A. Brown, A. Fedorov, I. Kokorin, Unexpected Scaling in Path Copying Trees, Association for Computing Machinery, 2023.","ama":"Aksenov V, Brown TA, Fedorov A, Kokorin I. <i>Unexpected Scaling in Path Copying Trees</i>. Association for Computing Machinery; 2023:438-440. doi:<a href=\"https://doi.org/10.1145/3572848.3577512\">10.1145/3572848.3577512</a>","ieee":"V. Aksenov, T. A. Brown, A. Fedorov, and I. Kokorin, <i>Unexpected scaling in path copying trees</i>. Association for Computing Machinery, 2023, pp. 438–440.","chicago":"Aksenov, Vitaly, Trevor A Brown, Alexander Fedorov, and Ilya Kokorin. <i>Unexpected Scaling in Path Copying Trees</i>. <i>Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming</i>. Association for Computing Machinery, 2023. <a href=\"https://doi.org/10.1145/3572848.3577512\">https://doi.org/10.1145/3572848.3577512</a>.","ista":"Aksenov V, Brown TA, Fedorov A, Kokorin I. 2023. Unexpected scaling in path copying trees, Association for Computing Machinery,p.","mla":"Aksenov, Vitaly, et al. “Unexpected Scaling in Path Copying Trees.” <i>Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming</i>, Association for Computing Machinery, 2023, pp. 438–40, doi:<a href=\"https://doi.org/10.1145/3572848.3577512\">10.1145/3572848.3577512</a>.","apa":"Aksenov, V., Brown, T. A., Fedorov, A., &#38; Kokorin, I. (2023). <i>Unexpected scaling in path copying trees</i>. <i>Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming</i> (pp. 438–440). Montreal, QB, Canada: Association for Computing Machinery. <a href=\"https://doi.org/10.1145/3572848.3577512\">https://doi.org/10.1145/3572848.3577512</a>"},"publisher":"Association for Computing Machinery"},{"citation":{"chicago":"Krumes, Alexandra, Adrian Vladu, Eldar Kurtic, Christoph Lampert, and Dan-Adrian Alistarh. “CrAM: A Compression-Aware Minimizer.” In <i>11th International Conference on Learning Representations </i>. OpenReview, 2023.","ista":"Krumes A, Vladu A, Kurtic E, Lampert C, Alistarh D-A. 2023. CrAM: A Compression-Aware Minimizer. 11th International Conference on Learning Representations . ICLR: International Conference on Learning Representations.","ieee":"A. Krumes, A. Vladu, E. Kurtic, C. Lampert, and D.-A. Alistarh, “CrAM: A Compression-Aware Minimizer,” in <i>11th International Conference on Learning Representations </i>, Kigali, Rwanda , 2023.","ama":"Krumes A, Vladu A, Kurtic E, Lampert C, Alistarh D-A. CrAM: A Compression-Aware Minimizer. In: <i>11th International Conference on Learning Representations </i>. OpenReview; 2023.","short":"A. Krumes, A. Vladu, E. Kurtic, C. Lampert, D.-A. Alistarh, in:, 11th International Conference on Learning Representations , OpenReview, 2023.","apa":"Krumes, A., Vladu, A., Kurtic, E., Lampert, C., &#38; Alistarh, D.-A. (2023). CrAM: A Compression-Aware Minimizer. In <i>11th International Conference on Learning Representations </i>. Kigali, Rwanda : OpenReview.","mla":"Krumes, Alexandra, et al. “CrAM: A Compression-Aware Minimizer.” <i>11th International Conference on Learning Representations </i>, OpenReview, 2023."},"publisher":"OpenReview","file":[{"date_created":"2024-07-22T09:09:45Z","file_name":"2023_ICLR_Peste.pdf","creator":"dernst","file_id":"17294","success":1,"date_updated":"2024-07-22T09:09:45Z","content_type":"application/pdf","access_level":"open_access","file_size":458201,"checksum":"a6eec897e13a91cdc3eeaf309801752c","relation":"main_file"}],"article_processing_charge":"No","main_file_link":[{"open_access":"1","url":"https://openreview.net/pdf?id=_eTZBs-yedr"}],"day":"01","date_published":"2023-05-01T00:00:00Z","file_date_updated":"2024-07-22T09:09:45Z","department":[{"_id":"GradSch"},{"_id":"DaAl"},{"_id":"ChLa"}],"related_material":{"link":[{"url":"https://github.com/IST-DASLab/CrAM","relation":"software"}],"record":[{"relation":"dissertation_contains","status":"public","id":"13074"}]},"acknowledgement":"AP, EK, DA received funding from the European Research Council (ERC) under the European\r\nUnion’s Horizon 2020 research and innovation programme (grant agreement No 805223 ScaleML). AV acknowledges the support of the French Agence Nationale de la Recherche (ANR), under grant ANR-21-CE48-0016 (project COMCOPT). We further acknowledge the support from the Scientific Service Units (SSU) of ISTA through resources provided by Scientific Computing (SciComp).","acknowledged_ssus":[{"_id":"ScienComp"}],"author":[{"last_name":"Peste","first_name":"Elena-Alexandra","full_name":"Peste, Elena-Alexandra","id":"32D78294-F248-11E8-B48F-1D18A9856A87"},{"last_name":"Vladu","first_name":"Adrian","full_name":"Vladu, Adrian"},{"full_name":"Kurtic, Eldar","first_name":"Eldar","last_name":"Kurtic","id":"47beb3a5-07b5-11eb-9b87-b108ec578218"},{"id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0001-8622-7887","full_name":"Lampert, Christoph","first_name":"Christoph","last_name":"Lampert"},{"last_name":"Alistarh","first_name":"Dan-Adrian","full_name":"Alistarh, Dan-Adrian","orcid":"0000-0003-3650-940X","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87"}],"ec_funded":1,"arxiv":1,"language":[{"iso":"eng"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","publication":"11th International Conference on Learning Representations ","ddc":["000"],"month":"05","quality_controlled":"1","external_id":{"arxiv":["2207.14200"]},"type":"conference","has_accepted_license":"1","year":"2023","project":[{"call_identifier":"H2020","_id":"268A44D6-B435-11E9-9278-68D0E5697425","name":"Elastic Coordination for Scalable Machine Learning","grant_number":"805223"}],"publication_status":"published","oa_version":"Published Version","conference":{"end_date":"2023-05-05","name":"ICLR: International Conference on Learning Representations","location":"Kigali, Rwanda ","start_date":"2023-05-01"},"date_updated":"2026-04-07T13:30:19Z","status":"public","_id":"13053","title":"CrAM: A Compression-Aware Minimizer","oa":1,"date_created":"2023-05-23T11:36:18Z","abstract":[{"text":"Deep neural networks (DNNs) often have to be compressed, via pruning and/or quantization, before they can be deployed in practical settings. In this work we propose a new compression-aware minimizer dubbed CrAM that modifies the optimization step in a principled way, in order to produce models whose local loss behavior is stable under compression operations such as pruning. Thus, dense models trained via CrAM should be compressible post-training, in a single step, without significant accuracy loss. Experimental results on standard benchmarks, such as residual networks for ImageNet classification and BERT models for language modelling, show that CrAM produces dense models that can be more accurate than the standard SGD/Adam-based baselines, but which are stable under weight pruning: specifically, we can prune models in one-shot to 70-80% sparsity with almost no accuracy loss, and to 90% with reasonable (∼1%) accuracy loss, which is competitive with gradual compression methods. Additionally, CrAM can produce sparse models which perform well for transfer learning, and it also works for semi-structured 2:4 pruning patterns supported by GPU hardware. The code for reproducing the results is available at this https URL .","lang":"eng"}],"corr_author":"1"},{"tmp":{"name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","image":"/images/cc_by.png","short":"CC BY (4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"ddc":["000"],"user_id":"317138e5-6ab7-11ef-aa6d-ffef3953e345","language":[{"iso":"eng"}],"publication":"Proceedings of the 35th ACM Symposium on Parallelism in Algorithms and Architectures","arxiv":1,"department":[{"_id":"DaAl"},{"_id":"GradSch"}],"author":[{"id":"2e711909-896a-11ed-bdf8-eb0f5a2984c6","full_name":"Fedorov, Alexander","last_name":"Fedorov","first_name":"Alexander"},{"id":"ed9595ea-2f8f-11ee-ba95-d2b546540783","full_name":"Hashemi, Diba","first_name":"Diba","last_name":"Hashemi"},{"first_name":"Giorgi","last_name":"Nadiradze","full_name":"Nadiradze, Giorgi","orcid":"0000-0001-5634-0731","id":"3279A00C-F248-11E8-B48F-1D18A9856A87"},{"id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-3650-940X","full_name":"Alistarh, Dan-Adrian","first_name":"Dan-Adrian","last_name":"Alistarh"}],"date_published":"2023-06-17T00:00:00Z","day":"17","article_processing_charge":"Yes (via OA deal)","file_date_updated":"2023-07-31T10:53:08Z","publisher":"Association for Computing Machinery","file":[{"date_created":"2023-07-31T10:53:08Z","creator":"dernst","file_name":"2023_SPAA_Fedorov.pdf","file_id":"13334","access_level":"open_access","file_size":2087937,"content_type":"application/pdf","date_updated":"2023-07-31T10:53:08Z","success":1,"checksum":"72e312aabf0c5248c99b5cd3a88e4c88","relation":"main_file"}],"citation":{"ama":"Fedorov A, Hashemi D, Nadiradze G, Alistarh D-A. Provably-efficient and internally-deterministic parallel Union-Find. In: <i>Proceedings of the 35th ACM Symposium on Parallelism in Algorithms and Architectures</i>. Association for Computing Machinery; 2023:261-271. doi:<a href=\"https://doi.org/10.1145/3558481.3591082\">10.1145/3558481.3591082</a>","ieee":"A. Fedorov, D. Hashemi, G. Nadiradze, and D.-A. Alistarh, “Provably-efficient and internally-deterministic parallel Union-Find,” in <i>Proceedings of the 35th ACM Symposium on Parallelism in Algorithms and Architectures</i>, Orlando, FL, United States, 2023, pp. 261–271.","chicago":"Fedorov, Alexander, Diba Hashemi, Giorgi Nadiradze, and Dan-Adrian Alistarh. “Provably-Efficient and Internally-Deterministic Parallel Union-Find.” In <i>Proceedings of the 35th ACM Symposium on Parallelism in Algorithms and Architectures</i>, 261–71. Association for Computing Machinery, 2023. <a href=\"https://doi.org/10.1145/3558481.3591082\">https://doi.org/10.1145/3558481.3591082</a>.","ista":"Fedorov A, Hashemi D, Nadiradze G, Alistarh D-A. 2023. Provably-efficient and internally-deterministic parallel Union-Find. Proceedings of the 35th ACM Symposium on Parallelism in Algorithms and Architectures. SPAA: Symposium on Parallelism in Algorithms and Architectures, 261–271.","short":"A. Fedorov, D. Hashemi, G. Nadiradze, D.-A. Alistarh, in:, Proceedings of the 35th ACM Symposium on Parallelism in Algorithms and Architectures, Association for Computing Machinery, 2023, pp. 261–271.","mla":"Fedorov, Alexander, et al. “Provably-Efficient and Internally-Deterministic Parallel Union-Find.” <i>Proceedings of the 35th ACM Symposium on Parallelism in Algorithms and Architectures</i>, Association for Computing Machinery, 2023, pp. 261–71, doi:<a href=\"https://doi.org/10.1145/3558481.3591082\">10.1145/3558481.3591082</a>.","apa":"Fedorov, A., Hashemi, D., Nadiradze, G., &#38; Alistarh, D.-A. (2023). Provably-efficient and internally-deterministic parallel Union-Find. In <i>Proceedings of the 35th ACM Symposium on Parallelism in Algorithms and Architectures</i> (pp. 261–271). Orlando, FL, United States: Association for Computing Machinery. <a href=\"https://doi.org/10.1145/3558481.3591082\">https://doi.org/10.1145/3558481.3591082</a>"},"doi":"10.1145/3558481.3591082","has_accepted_license":"1","type":"conference","external_id":{"isi":["001108889000024"],"arxiv":["2304.09331"]},"quality_controlled":"1","month":"06","publication_identifier":{"isbn":["9781450395458"]},"conference":{"location":"Orlando, FL, United States","start_date":"2023-06-17","name":"SPAA: Symposium on Parallelism in Algorithms and Architectures","end_date":"2023-06-19"},"date_updated":"2025-09-09T12:43:18Z","oa_version":"Published Version","scopus_import":"1","publication_status":"published","page":"261-271","year":"2023","corr_author":"1","isi":1,"abstract":[{"lang":"eng","text":"Determining the degree of inherent parallelism in classical sequential algorithms and leveraging it for fast parallel execution is a key topic in parallel computing, and detailed analyses are known for a wide range of classical algorithms. In this paper, we perform the first such analysis for the fundamental Union-Find problem, in which we are given a graph as a sequence of edges, and must maintain its connectivity structure under edge additions. We prove that classic sequential algorithms for this problem are well-parallelizable under reasonable assumptions, addressing a conjecture by [Blelloch, 2017]. More precisely, we show via a new potential argument that, under uniform random edge ordering, parallel union-find operations are unlikely to interfere: T concurrent threads processing the graph in parallel will encounter memory contention O(T2 · log |V| · log |E|) times in expectation, where |E| and |V| are the number of edges and nodes in the graph, respectively. We leverage this result to design a new parallel Union-Find algorithm that is both internally deterministic, i.e., its results are guaranteed to match those of a sequential execution, but also work-efficient and scalable, as long as the number of threads T is O(|E|1 over 3 - ε), for an arbitrarily small constant ε > 0, which holds for most large real-world graphs. We present lower bounds which show that our analysis is close to optimal, and experimental results suggesting that the performance cost of internal determinism is limited."}],"oa":1,"date_created":"2023-07-23T22:01:12Z","_id":"13262","status":"public","title":"Provably-efficient and internally-deterministic parallel Union-Find"},{"isi":1,"date_created":"2023-09-24T22:01:11Z","oa":1,"abstract":[{"lang":"eng","text":"We introduce extension-based proofs, a class of impossibility proofs that includes valency arguments. They are modelled as an interaction between a prover and a protocol. Using proofs based on combinatorial topology, it has been shown that it is impossible to deterministically solve -set agreement among  processes or approximate agreement on a cycle of length 4 among  processes in a wait-free manner in asynchronous models where processes communicate using objects that can be constructed from shared registers. However, it was unknown whether proofs based on simpler techniques were possible. We show that these impossibility results cannot be obtained by extension-based proofs in the iterated snapshot model and, hence, extension-based proofs are limited in power."}],"status":"public","_id":"14364","title":"Why extension-based proofs fail","date_updated":"2025-05-14T11:26:06Z","oa_version":"Preprint","scopus_import":"1","page":"913-944","publication_status":"published","year":"2023","project":[{"call_identifier":"H2020","_id":"268A44D6-B435-11E9-9278-68D0E5697425","grant_number":"805223","name":"Elastic Coordination for Scalable Machine Learning"}],"volume":52,"doi":"10.1137/20M1375851","external_id":{"isi":["001082972300004"],"arxiv":["1811.01421"]},"type":"journal_article","article_type":"original","month":"07","quality_controlled":"1","publication_identifier":{"issn":["0097-5397"],"eissn":["1095-7111"]},"arxiv":1,"publication":"SIAM Journal on Computing","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","language":[{"iso":"eng"}],"department":[{"_id":"DaAl"}],"related_material":{"record":[{"relation":"earlier_version","status":"public","id":"6676"}]},"acknowledgement":"We would like to thank Valerie King, Toniann Pitassi, and Michael Saks for helpful discussions and Shi Hao Liu for his useful feedback.\r\nThis research was supported by the Natural Science and Engineering Research Council of Canada under grants RGPIN-2015-05080 and RGPIN-2020-04178, a postgraduate scholarship, and a postdoctoral fellowship; a University of Toronto postdoctoral fellowship; the National Science Foundation under grants CCF-1217921, CCF-1301926, CCF-1637385, CCF-1650596, and IIS-1447786; the U.S. Department of Energy under grant ER26116/DE-SC0008923; the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme grant agreement 805223 ScaleML; and the Oracle and Intel corporations. Some of the work on this paper was done while Faith Ellen was visiting IST Austria.","author":[{"orcid":"0000-0003-3650-940X","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","first_name":"Dan-Adrian","last_name":"Alistarh","full_name":"Alistarh, Dan-Adrian"},{"full_name":"Aspnes, James","first_name":"James","last_name":"Aspnes"},{"full_name":"Ellen, Faith","first_name":"Faith","last_name":"Ellen"},{"first_name":"Rati","last_name":"Gelashvili","full_name":"Gelashvili, Rati"},{"first_name":"Leqi","last_name":"Zhu","full_name":"Zhu, Leqi","id":"a2117c59-cee4-11ed-b9d0-874ecf0f8ac5"}],"ec_funded":1,"article_processing_charge":"No","date_published":"2023-07-25T00:00:00Z","main_file_link":[{"open_access":"1","url":"https://arxiv.org/abs/1811.01421"}],"day":"25","intvolume":"        52","issue":"4","citation":{"mla":"Alistarh, Dan-Adrian, et al. “Why Extension-Based Proofs Fail.” <i>SIAM Journal on Computing</i>, vol. 52, no. 4, Society for Industrial and Applied Mathematics, 2023, pp. 913–44, doi:<a href=\"https://doi.org/10.1137/20M1375851\">10.1137/20M1375851</a>.","apa":"Alistarh, D.-A., Aspnes, J., Ellen, F., Gelashvili, R., &#38; Zhu, L. (2023). Why extension-based proofs fail. <i>SIAM Journal on Computing</i>. Society for Industrial and Applied Mathematics. <a href=\"https://doi.org/10.1137/20M1375851\">https://doi.org/10.1137/20M1375851</a>","short":"D.-A. Alistarh, J. Aspnes, F. Ellen, R. Gelashvili, L. Zhu, SIAM Journal on Computing 52 (2023) 913–944.","ama":"Alistarh D-A, Aspnes J, Ellen F, Gelashvili R, Zhu L. Why extension-based proofs fail. <i>SIAM Journal on Computing</i>. 2023;52(4):913-944. doi:<a href=\"https://doi.org/10.1137/20M1375851\">10.1137/20M1375851</a>","ista":"Alistarh D-A, Aspnes J, Ellen F, Gelashvili R, Zhu L. 2023. Why extension-based proofs fail. SIAM Journal on Computing. 52(4), 913–944.","chicago":"Alistarh, Dan-Adrian, James Aspnes, Faith Ellen, Rati Gelashvili, and Leqi Zhu. “Why Extension-Based Proofs Fail.” <i>SIAM Journal on Computing</i>. Society for Industrial and Applied Mathematics, 2023. <a href=\"https://doi.org/10.1137/20M1375851\">https://doi.org/10.1137/20M1375851</a>.","ieee":"D.-A. Alistarh, J. Aspnes, F. Ellen, R. Gelashvili, and L. Zhu, “Why extension-based proofs fail,” <i>SIAM Journal on Computing</i>, vol. 52, no. 4. Society for Industrial and Applied Mathematics, pp. 913–944, 2023."},"publisher":"Society for Industrial and Applied Mathematics"},{"alternative_title":["PMLR"],"publication_identifier":{"eissn":["2640-3498"]},"month":"07","quality_controlled":"1","external_id":{"arxiv":["2302.04852"]},"type":"conference","volume":202,"citation":{"ama":"Nikdan M, Pegolotti T, Iofinova EB, Kurtic E, Alistarh D-A. SparseProp: Efficient sparse backpropagation for faster training of neural networks at the edge. In: <i>Proceedings of the 40th International Conference on Machine Learning</i>. Vol 202. ML Research Press; 2023:26215-26227.","ieee":"M. Nikdan, T. Pegolotti, E. B. Iofinova, E. Kurtic, and D.-A. Alistarh, “SparseProp: Efficient sparse backpropagation for faster training of neural networks at the edge,” in <i>Proceedings of the 40th International Conference on Machine Learning</i>, Honolulu, Hawaii, HI, United States, 2023, vol. 202, pp. 26215–26227.","chicago":"Nikdan, Mahdi, Tommaso Pegolotti, Eugenia B Iofinova, Eldar Kurtic, and Dan-Adrian Alistarh. “SparseProp: Efficient Sparse Backpropagation for Faster Training of Neural Networks at the Edge.” In <i>Proceedings of the 40th International Conference on Machine Learning</i>, 202:26215–27. ML Research Press, 2023.","ista":"Nikdan M, Pegolotti T, Iofinova EB, Kurtic E, Alistarh D-A. 2023. SparseProp: Efficient sparse backpropagation for faster training of neural networks at the edge. Proceedings of the 40th International Conference on Machine Learning. ICML: International Conference on Machine Learning, PMLR, vol. 202, 26215–26227.","short":"M. Nikdan, T. Pegolotti, E.B. Iofinova, E. Kurtic, D.-A. Alistarh, in:, Proceedings of the 40th International Conference on Machine Learning, ML Research Press, 2023, pp. 26215–26227.","mla":"Nikdan, Mahdi, et al. “SparseProp: Efficient Sparse Backpropagation for Faster Training of Neural Networks at the Edge.” <i>Proceedings of the 40th International Conference on Machine Learning</i>, vol. 202, ML Research Press, 2023, pp. 26215–27.","apa":"Nikdan, M., Pegolotti, T., Iofinova, E. B., Kurtic, E., &#38; Alistarh, D.-A. (2023). SparseProp: Efficient sparse backpropagation for faster training of neural networks at the edge. In <i>Proceedings of the 40th International Conference on Machine Learning</i> (Vol. 202, pp. 26215–26227). Honolulu, Hawaii, HI, United States: ML Research Press."},"publisher":"ML Research Press","intvolume":"       202","article_processing_charge":"No","main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2302.04852"}],"day":"30","date_published":"2023-07-30T00:00:00Z","ec_funded":1,"author":[{"id":"66374281-f394-11eb-9cf6-869147deecc0","first_name":"Mahdi","last_name":"Nikdan","full_name":"Nikdan, Mahdi"},{"full_name":"Pegolotti, Tommaso","first_name":"Tommaso","last_name":"Pegolotti"},{"first_name":"Eugenia B","last_name":"Iofinova","full_name":"Iofinova, Eugenia B","orcid":"0000-0002-7778-3221","id":"f9a17499-f6e0-11ea-865d-fdf9a3f77117"},{"last_name":"Kurtic","first_name":"Eldar","full_name":"Kurtic, Eldar","id":"47beb3a5-07b5-11eb-9b87-b108ec578218"},{"full_name":"Alistarh, Dan-Adrian","first_name":"Dan-Adrian","last_name":"Alistarh","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-3650-940X"}],"department":[{"_id":"DaAl"}],"acknowledgement":"We would like to thank Elias Frantar for his valuable assistance and support at the outset of this project, and the anonymous ICML and SNN reviewers for very constructive feedback. EI was supported in part by the FWF DK VGSCO, grant agreement number W1260-N35. DA acknowledges generous ERC support, via Starting Grant 805223 ScaleML. ","arxiv":1,"publication":"Proceedings of the 40th International Conference on Machine Learning","language":[{"iso":"eng"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","title":"SparseProp: Efficient sparse backpropagation for faster training of neural networks at the edge","status":"public","_id":"14460","date_created":"2023-10-29T23:01:17Z","oa":1,"abstract":[{"text":"We provide an efficient implementation of the backpropagation algorithm, specialized to the case where the weights of the neural network being trained are sparse. Our algorithm is general, as it applies to arbitrary (unstructured) sparsity and common layer types (e.g., convolutional or linear). We provide a fast vectorized implementation on commodity CPUs, and show that it can yield speedups in end-to-end runtime experiments, both in transfer learning using already-sparsified networks, and in training sparse networks from scratch. Thus, our results provide the first support for sparse training on commodity hardware.","lang":"eng"}],"corr_author":"1","project":[{"_id":"268A44D6-B435-11E9-9278-68D0E5697425","call_identifier":"H2020","name":"Elastic Coordination for Scalable Machine Learning","grant_number":"805223"}],"year":"2023","publication_status":"published","page":"26215-26227","oa_version":"Preprint","scopus_import":"1","date_updated":"2025-04-14T07:49:12Z","conference":{"start_date":"2023-07-23","location":"Honolulu, Hawaii, HI, United States","end_date":"2023-07-29","name":"ICML: International Conference on Machine Learning"}},{"publication":"Proceedings of the 40th International Conference on Machine Learning","language":[{"iso":"eng"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","arxiv":1,"acknowledgement":"The authors gratefully acknowledge funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No 805223 ScaleML), as well as experimental support from the IST Austria IT department, in particular Stefano Elefante, Andrei Hornoiu, and Alois Schloegl. AV acknowledges the support of the French Agence Nationale de la Recherche (ANR), under grant ANR-21-CE48-0016 (project COMCOPT), the support of Fondation Hadamard with a PRMO grant, and the support of CNRS with a CoopIntEER IEA grant (project ALFRED).","related_material":{"record":[{"id":"17490","relation":"dissertation_contains","status":"public"}]},"department":[{"_id":"DaAl"}],"ec_funded":1,"author":[{"id":"D0CF4148-C985-11E9-8066-0BDEE5697425","first_name":"Ilia","last_name":"Markov","full_name":"Markov, Ilia"},{"last_name":"Vladu","first_name":"Adrian","full_name":"Vladu, Adrian"},{"full_name":"Guo, Qi","last_name":"Guo","first_name":"Qi"},{"full_name":"Alistarh, Dan-Adrian","last_name":"Alistarh","first_name":"Dan-Adrian","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-3650-940X"}],"acknowledged_ssus":[{"_id":"ScienComp"}],"date_published":"2023-07-30T00:00:00Z","main_file_link":[{"url":"https://doi.org/10.48550/arXiv.2302.02390","open_access":"1"}],"day":"30","article_processing_charge":"No","intvolume":"       202","publisher":"ML Research Press","citation":{"mla":"Markov, Ilia, et al. “Quantized Distributed Training of Large Models with Convergence Guarantees.” <i>Proceedings of the 40th International Conference on Machine Learning</i>, vol. 202, ML Research Press, 2023, pp. 24020–44.","apa":"Markov, I., Vladu, A., Guo, Q., &#38; Alistarh, D.-A. (2023). Quantized distributed training of large models with convergence guarantees. In <i>Proceedings of the 40th International Conference on Machine Learning</i> (Vol. 202, pp. 24020–24044). Honolulu, Hawaii, HI, United States: ML Research Press.","short":"I. Markov, A. Vladu, Q. Guo, D.-A. Alistarh, in:, Proceedings of the 40th International Conference on Machine Learning, ML Research Press, 2023, pp. 24020–24044.","ama":"Markov I, Vladu A, Guo Q, Alistarh D-A. Quantized distributed training of large models with convergence guarantees. In: <i>Proceedings of the 40th International Conference on Machine Learning</i>. Vol 202. ML Research Press; 2023:24020-24044.","ieee":"I. Markov, A. Vladu, Q. Guo, and D.-A. Alistarh, “Quantized distributed training of large models with convergence guarantees,” in <i>Proceedings of the 40th International Conference on Machine Learning</i>, Honolulu, Hawaii, HI, United States, 2023, vol. 202, pp. 24020–24044.","chicago":"Markov, Ilia, Adrian Vladu, Qi Guo, and Dan-Adrian Alistarh. “Quantized Distributed Training of Large Models with Convergence Guarantees.” In <i>Proceedings of the 40th International Conference on Machine Learning</i>, 202:24020–44. ML Research Press, 2023.","ista":"Markov I, Vladu A, Guo Q, Alistarh D-A. 2023. Quantized distributed training of large models with convergence guarantees. Proceedings of the 40th International Conference on Machine Learning. ICML: International Conference on Machine Learning, PMLR, vol. 202, 24020–24044."},"volume":202,"type":"conference","external_id":{"arxiv":["2302.02390"]},"quality_controlled":"1","month":"07","alternative_title":["PMLR"],"publication_identifier":{"eissn":["2640-3498"]},"conference":{"start_date":"2023-07-23","location":"Honolulu, Hawaii, HI, United States","name":"ICML: International Conference on Machine Learning","end_date":"2023-07-29"},"date_updated":"2026-04-07T13:00:54Z","scopus_import":"1","oa_version":"Preprint","page":"24020-24044","publication_status":"published","year":"2023","project":[{"_id":"268A44D6-B435-11E9-9278-68D0E5697425","call_identifier":"H2020","grant_number":"805223","name":"Elastic Coordination for Scalable Machine Learning"}],"corr_author":"1","abstract":[{"lang":"eng","text":"Communication-reduction techniques are a popular way to improve scalability in data-parallel training of deep neural networks (DNNs). The recent emergence of large language models such as GPT has created the need for new approaches to exploit data-parallelism. Among these, fully-sharded data parallel (FSDP) training is highly popular, yet it still encounters scalability bottlenecks. One reason is that applying compression techniques to FSDP is challenging: as the vast majority of the communication involves the model’s weights, direct compression alters convergence and leads to accuracy loss. We present QSDP, a variant of FSDP which supports both gradient and weight quantization with theoretical guarantees, is simple to implement and has essentially no overheads. To derive QSDP we prove that a natural modification of SGD achieves convergence even when we only maintain quantized weights, and thus the domain over which we train consists of quantized points and is, therefore, highly non-convex. We validate this approach by training GPT-family models with up to 1.3 billion parameters on a multi-node cluster. Experiments show that QSDP preserves model accuracy, while completely removing the communication bottlenecks of FSDP, providing end-to-end speedups of up to 2.2x."}],"date_created":"2023-10-29T23:01:17Z","oa":1,"status":"public","_id":"14461","title":"Quantized distributed training of large models with convergence guarantees"},{"abstract":[{"lang":"eng","text":"In the last few years, various communication compression techniques have emerged as an indispensable tool helping to alleviate the communication bottleneck in distributed learning. However, despite the fact biased compressors often show superior performance in practice when compared to the much more studied and understood unbiased compressors, very little is known about them. In this work we study three classes of biased compression operators, two of which are new, and their performance when applied to (stochastic) gradient descent and distributed (stochastic) gradient descent. We show for the first time that biased compressors can lead to linear convergence rates both in the single node and distributed settings. We prove that distributed compressed SGD method, employed with error feedback mechanism, enjoys the ergodic rate O(δLexp[−μKδL]+(C+δD)Kμ), where δ≥1 is a compression parameter which grows when more compression is applied, L and μ are the smoothness and strong convexity constants, C captures stochastic gradient noise (C=0 if full gradients are computed on each node) and D captures the variance of the gradients at the optimum (D=0 for over-parameterized models). Further, via a theoretical study of several synthetic and empirical distributions of communicated gradients, we shed light on why and by how much biased compressors outperform their unbiased variants. Finally, we propose several new biased compressors with promising theoretical guarantees and practical performance."}],"oa":1,"date_created":"2024-01-16T12:13:36Z","_id":"14815","status":"public","title":"On biased compression for distributed learning","corr_author":"1","isi":1,"publication_status":"published","page":"1-50","year":"2023","date_updated":"2024-10-09T21:07:52Z","oa_version":"Published Version","article_type":"original","quality_controlled":"1","month":"10","publication_identifier":{"eissn":["1533-7928"]},"volume":24,"has_accepted_license":"1","type":"journal_article","external_id":{"arxiv":["2002.12410"],"isi":["001111578500001"]},"day":"01","date_published":"2023-10-01T00:00:00Z","article_processing_charge":"Yes (in subscription journal)","file_date_updated":"2024-01-16T12:13:27Z","intvolume":"        24","file":[{"file_id":"14816","creator":"dernst","file_name":"2023_JMLR_Beznosikov.pdf","date_created":"2024-01-16T12:13:27Z","checksum":"c50f2b9db53938b755e30a085f464059","relation":"main_file","file_size":1510993,"access_level":"open_access","date_updated":"2024-01-16T12:13:27Z","content_type":"application/pdf","success":1}],"publisher":"Journal of Machine Learning Research","citation":{"mla":"Beznosikov, Aleksandr, et al. “On Biased Compression for Distributed Learning.” <i>Journal of Machine Learning Research</i>, vol. 24, Journal of Machine Learning Research, 2023, pp. 1–50.","apa":"Beznosikov, A., Horvath, S., Richtarik, P., &#38; Safaryan, M. (2023). On biased compression for distributed learning. <i>Journal of Machine Learning Research</i>. Journal of Machine Learning Research.","ama":"Beznosikov A, Horvath S, Richtarik P, Safaryan M. On biased compression for distributed learning. <i>Journal of Machine Learning Research</i>. 2023;24:1-50.","chicago":"Beznosikov, Aleksandr, Samuel Horvath, Peter Richtarik, and Mher Safaryan. “On Biased Compression for Distributed Learning.” <i>Journal of Machine Learning Research</i>. Journal of Machine Learning Research, 2023.","ieee":"A. Beznosikov, S. Horvath, P. Richtarik, and M. Safaryan, “On biased compression for distributed learning,” <i>Journal of Machine Learning Research</i>, vol. 24. Journal of Machine Learning Research, pp. 1–50, 2023.","ista":"Beznosikov A, Horvath S, Richtarik P, Safaryan M. 2023. On biased compression for distributed learning. Journal of Machine Learning Research. 24, 1–50.","short":"A. Beznosikov, S. Horvath, P. Richtarik, M. Safaryan, Journal of Machine Learning Research 24 (2023) 1–50."},"tmp":{"name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","image":"/images/cc_by.png","short":"CC BY (4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"ddc":["000"],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","publication":"Journal of Machine Learning Research","language":[{"iso":"eng"}],"arxiv":1,"acknowledgement":"The work in Sections 1-5 was conducted while A. Beznosikov was a research intern in the Optimizationand Machine Learning Lab of Peter Richtárik at KAUST; this visit was funded by the KAUST Baseline Research Funding Scheme. The work of A. Beznosikov in Section 6 was conducted in Skoltech and was supported by Ministry of Science and Higher Education grant No. 075-10-2021-068. ","department":[{"_id":"DaAl"}],"author":[{"last_name":"Beznosikov","first_name":"Aleksandr","full_name":"Beznosikov, Aleksandr"},{"first_name":"Samuel","last_name":"Horvath","full_name":"Horvath, Samuel"},{"full_name":"Richtarik, Peter","last_name":"Richtarik","first_name":"Peter"},{"id":"dd546b39-0804-11ed-9c55-ef075c39778d","full_name":"Safaryan, Mher","first_name":"Mher","last_name":"Safaryan"}]},{"_id":"13179","status":"public","title":"CQS: A formally-verified framework for fair and abortable synchronization","date_created":"2023-07-02T22:00:43Z","oa":1,"abstract":[{"text":"Writing concurrent code that is both correct and efficient is notoriously difficult. Thus, programmers often prefer to use synchronization abstractions, which render code simpler and easier to reason about. Despite a wealth of work on this topic, there is still a gap between the rich semantics provided by synchronization abstractions in modern programming languages—specifically, fair FIFO ordering of synchronization requests and support for abortable operations—and frameworks for implementing it correctly and efficiently. Supporting such semantics is critical given the rising popularity of constructs for asynchronous programming, such as coroutines, which abort frequently and are cheaper to suspend and resume compared to native threads.\r\n\r\nThis paper introduces a new framework called CancellableQueueSynchronizer (CQS), which enables simple yet efficient implementations of a wide range of fair and abortable synchronization primitives: mutexes, semaphores, barriers, count-down latches, and blocking pools. Our main contribution is algorithmic, as implementing both fairness and abortability efficiently at this level of generality is non-trivial. Importantly, all our algorithms, including the CQS framework and the primitives built on top of it, come with formal proofs in the Iris framework for Coq for many of their properties. These proofs are modular, so it is easy to show correctness for new primitives implemented on top of CQS. From a practical perspective, implementation of CQS for native threads on the JVM improves throughput by up to two orders of magnitude over Java’s AbstractQueuedSynchronizer, the only practical abstraction offering similar semantics. Further, we successfully integrated CQS as a core component of the popular Kotlin Coroutines library, validating the framework’s practical impact and expressiveness in a real-world environment. In sum, CancellableQueueSynchronizer is the first framework to combine expressiveness with formal guarantees and solid practical performance. Our approach should be extensible to other languages and families of synchronization primitives.","lang":"eng"}],"das_tickbox":"1","corr_author":"1","year":"2023","publication_status":"published","scopus_import":"1","oa_version":"Published Version","date_updated":"2026-07-06T12:12:08Z","publication_identifier":{"eissn":["2475-1421"]},"article_type":"original","month":"06","quality_controlled":"1","type":"journal_article","has_accepted_license":"1","volume":7,"doi":"10.1145/3591230","intvolume":"         7","citation":{"mla":"Koval, Nikita, et al. “CQS: A Formally-Verified Framework for Fair and Abortable Synchronization.” <i>Proceedings of the ACM on Programming Languages</i>, vol. 7, 116, Association for Computing Machinery, 2023, doi:<a href=\"https://doi.org/10.1145/3591230\">10.1145/3591230</a>.","apa":"Koval, N., Khalanskiy, D., &#38; Alistarh, D.-A. (2023). CQS: A formally-verified framework for fair and abortable synchronization. <i>Proceedings of the ACM on Programming Languages</i>. Association for Computing Machinery. <a href=\"https://doi.org/10.1145/3591230\">https://doi.org/10.1145/3591230</a>","ama":"Koval N, Khalanskiy D, Alistarh D-A. CQS: A formally-verified framework for fair and abortable synchronization. <i>Proceedings of the ACM on Programming Languages</i>. 2023;7. doi:<a href=\"https://doi.org/10.1145/3591230\">10.1145/3591230</a>","ista":"Koval N, Khalanskiy D, Alistarh D-A. 2023. CQS: A formally-verified framework for fair and abortable synchronization. Proceedings of the ACM on Programming Languages. 7, 116.","ieee":"N. Koval, D. Khalanskiy, and D.-A. Alistarh, “CQS: A formally-verified framework for fair and abortable synchronization,” <i>Proceedings of the ACM on Programming Languages</i>, vol. 7. Association for Computing Machinery, 2023.","chicago":"Koval, Nikita, Dmitry Khalanskiy, and Dan-Adrian Alistarh. “CQS: A Formally-Verified Framework for Fair and Abortable Synchronization.” <i>Proceedings of the ACM on Programming Languages</i>. Association for Computing Machinery, 2023. <a href=\"https://doi.org/10.1145/3591230\">https://doi.org/10.1145/3591230</a>.","short":"N. Koval, D. Khalanskiy, D.-A. Alistarh, Proceedings of the ACM on Programming Languages 7 (2023)."},"publisher":"Association for Computing Machinery","file":[{"date_created":"2023-07-03T13:09:39Z","file_name":"2023_ACMProgram.Lang._Koval.pdf","creator":"alisjak","file_id":"13187","access_level":"open_access","file_size":1266773,"content_type":"application/pdf","date_updated":"2023-07-03T13:09:39Z","success":1,"relation":"main_file","checksum":"5dba6e73f0ed79adbdae14d165bc2f68"}],"article_processing_charge":"No","date_published":"2023-06-06T00:00:00Z","day":"06","file_date_updated":"2023-07-03T13:09:39Z","article_number":"116","department":[{"_id":"DaAl"}],"author":[{"id":"2F4DB10C-F248-11E8-B48F-1D18A9856A87","full_name":"Koval, Nikita","last_name":"Koval","first_name":"Nikita"},{"first_name":"Dmitry","last_name":"Khalanskiy","full_name":"Khalanskiy, Dmitry"},{"orcid":"0000-0003-3650-940X","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","first_name":"Dan-Adrian","last_name":"Alistarh","full_name":"Alistarh, Dan-Adrian"}],"tmp":{"name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","image":"/images/cc_by.png","short":"CC BY (4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"ddc":["000"],"publication":"Proceedings of the ACM on Programming Languages","language":[{"iso":"eng"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87"},{"abstract":[{"text":"The sinkless orientation problem plays a key role in understanding the foundations of distributed computing. The problem can be used to separate two fundamental models of distributed graph algorithms, LOCAL and SLOCAL: the locality of sinkless orientation is Ω(log n) in the deterministic LOCAL model and O(log log n) in the deterministic SLOCAL model. Both of these results are known by prior work, but here we give new simple, self-contained proofs for them.","lang":"eng"}],"oa":1,"date_created":"2025-07-10T13:11:47Z","status":"public","_id":"19983","title":"Sinkless Orientation Made Simple","das_tickbox":"1","page":"175-191","publication_status":"published","OA_type":"green","year":"2023","project":[{"name":"Elastic Coordination for Scalable Machine Learning","grant_number":"805223","call_identifier":"H2020","_id":"268A44D6-B435-11E9-9278-68D0E5697425"},{"name":"Fast Algorithms for a Reactive Network Layer","grant_number":"P33775","_id":"bd9e3a2e-d553-11ed-ba76-8aa684ce17fe"}],"conference":{"name":"SOSA: Symposium on Simplicity in Algorithms","end_date":"2023-01-25","location":"Florence, Italy","start_date":"2023-01-23"},"date_updated":"2026-07-06T11:57:38Z","oa_version":"Preprint","quality_controlled":"1","month":"01","publication_identifier":{"eisbn":["9781611977585"]},"doi":"10.1137/1.9781611977585.ch17","type":"book_chapter","external_id":{"arxiv":["2108.02655"]},"date_published":"2023-01-12T00:00:00Z","main_file_link":[{"url":"https://doi.org/10.48550/arXiv.2108.02655","open_access":"1"}],"day":"12","article_processing_charge":"No","publisher":"Society for Industrial and Applied Mathematics","citation":{"short":"A. Balliu, J. Korhonen, F. Kuhn, H. Lievonen, D. Olivetti, S. Pai, A. Paz, J. Rybicki, S. Schmid, J. Studený, J. Suomela, J. Uitto, in:, Symposium on Simplicity in Algorithms, Society for Industrial and Applied Mathematics, 2023, pp. 175–191.","chicago":"Balliu, Alkida, Janne Korhonen, Fabian Kuhn, Henrik Lievonen, Dennis Olivetti, Shreyas Pai, Ami Paz, et al. “Sinkless Orientation Made Simple.” In <i>Symposium on Simplicity in Algorithms</i>, 175–91. Society for Industrial and Applied Mathematics, 2023. <a href=\"https://doi.org/10.1137/1.9781611977585.ch17\">https://doi.org/10.1137/1.9781611977585.ch17</a>.","ista":"Balliu A, Korhonen J, Kuhn F, Lievonen H, Olivetti D, Pai S, Paz A, Rybicki J, Schmid S, Studený J, Suomela J, Uitto J. 2023.Sinkless Orientation Made Simple. In: Symposium on Simplicity in Algorithms. , 175–191.","ieee":"A. Balliu <i>et al.</i>, “Sinkless Orientation Made Simple,” in <i>Symposium on Simplicity in Algorithms</i>, Society for Industrial and Applied Mathematics, 2023, pp. 175–191.","ama":"Balliu A, Korhonen J, Kuhn F, et al. Sinkless Orientation Made Simple. In: <i>Symposium on Simplicity in Algorithms</i>. Society for Industrial and Applied Mathematics; 2023:175-191. doi:<a href=\"https://doi.org/10.1137/1.9781611977585.ch17\">10.1137/1.9781611977585.ch17</a>","apa":"Balliu, A., Korhonen, J., Kuhn, F., Lievonen, H., Olivetti, D., Pai, S., … Uitto, J. (2023). Sinkless Orientation Made Simple. In <i>Symposium on Simplicity in Algorithms</i> (pp. 175–191). Florence, Italy: Society for Industrial and Applied Mathematics. <a href=\"https://doi.org/10.1137/1.9781611977585.ch17\">https://doi.org/10.1137/1.9781611977585.ch17</a>","mla":"Balliu, Alkida, et al. “Sinkless Orientation Made Simple.” <i>Symposium on Simplicity in Algorithms</i>, Society for Industrial and Applied Mathematics, 2023, pp. 175–91, doi:<a href=\"https://doi.org/10.1137/1.9781611977585.ch17\">10.1137/1.9781611977585.ch17</a>."},"publication":"Symposium on Simplicity in Algorithms","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","language":[{"iso":"eng"}],"arxiv":1,"acknowledgement":"We thank the anonymous reviewers for their helpful feedback on previous versions of this work. Parts ofthis work appeared in DISC 2021 as a brief announcement [ 21]. This work was supported in part by theEuropean Research Council (ERC) under the European Union’s Horizon 2020 research and innovationprogramme (grant agreement No 805223 ScaleML), the Academy of Finland (grant agreement No 333837),the Austrian Science Fund (FWF) and netIDEE (grant agreement No P 33775-N), and the AustrianScience Fund (FWF) project DELTA (grant agreement No I 5025-N).","department":[{"_id":"DaAl"}],"author":[{"full_name":"Balliu, Alkida","first_name":"Alkida","last_name":"Balliu"},{"full_name":"Korhonen, Janne","first_name":"Janne","last_name":"Korhonen","id":"C5402D42-15BC-11E9-A202-CA2BE6697425"},{"last_name":"Kuhn","first_name":"Fabian","full_name":"Kuhn, Fabian"},{"last_name":"Lievonen","first_name":"Henrik","full_name":"Lievonen, Henrik"},{"full_name":"Olivetti, Dennis","first_name":"Dennis","last_name":"Olivetti"},{"first_name":"Shreyas","last_name":"Pai","full_name":"Pai, Shreyas"},{"last_name":"Paz","first_name":"Ami","full_name":"Paz, Ami"},{"orcid":"0000-0002-6432-6646","id":"334EFD2E-F248-11E8-B48F-1D18A9856A87","last_name":"Rybicki","first_name":"Joel","full_name":"Rybicki, Joel"},{"last_name":"Schmid","first_name":"Stefan","full_name":"Schmid, Stefan"},{"last_name":"Studený","first_name":"Jan","full_name":"Studený, Jan"},{"last_name":"Suomela","first_name":"Jukka","full_name":"Suomela, Jukka"},{"full_name":"Uitto, Jara","last_name":"Uitto","first_name":"Jara"}],"ec_funded":1,"OA_place":"repository"},{"date_created":"2024-05-05T22:01:04Z","oa":1,"abstract":[{"text":"Knowledge distillation is a popular approach for enhancing the performance of \"student\" models, with lower representational capacity, by taking advantage of more powerful \"teacher\" models. Despite its apparent simplicity, the underlying mechanics behind knowledge distillation (KD) are not yet fully understood. In this work, we shed new light on the inner workings of this method, by examining it from an optimization perspective. Specifically, we show that, in the context of linear and deep linear models, KD can be interpreted as a novel type of stochastic variance reduction mechanism. We provide a detailed convergence analysis of the resulting dynamics, which hold under standard assumptions for both strongly-convex and non-convex losses, showing that KD acts as a form of \\emph{partial variance reduction}, which can reduce the stochastic gradient noise, but may not eliminate it completely, depending on the properties of the teacher'' model. Our analysis puts further emphasis on the need for careful parametrization of KD, in particular w.r.t. the weighting of the distillation loss, and is validated empirically on both linear models and deep neural networks.","lang":"eng"}],"status":"public","_id":"15363","title":"Knowledge distillation performs partial variance reduction","das_tickbox":"1","corr_author":"1","publication_status":"published","year":"2023","project":[{"call_identifier":"H2020","_id":"fc2ed2f7-9c52-11eb-aca3-c01059dda49c","name":"IST-BRIDGE: International postdoctoral program","grant_number":"101034413"}],"conference":{"name":"NeurIPS: Neural Information Processing Systems","end_date":"2023-12-16","start_date":"2023-12-10","location":"New Orleans, LA, United States"},"date_updated":"2026-07-07T06:37:28Z","oa_version":"Published Version","scopus_import":"1","month":"12","quality_controlled":"1","alternative_title":["Advances in Neural Information Processing Systems"],"publication_identifier":{"issn":["1049-5258"]},"has_accepted_license":"1","volume":36,"external_id":{"arxiv":["2305.17581"]},"type":"conference","article_processing_charge":"Yes","date_published":"2023-12-15T00:00:00Z","day":"15","file_date_updated":"2024-05-22T08:08:08Z","intvolume":"        36","citation":{"short":"M. Safaryan, A. Krumes, D.-A. Alistarh, in:, 36th Conference on Neural Information Processing Systems, Neural Information Processing Systems Foundation, 2023.","ama":"Safaryan M, Krumes A, Alistarh D-A. Knowledge distillation performs partial variance reduction. In: <i>36th Conference on Neural Information Processing Systems</i>. Vol 36. Neural Information Processing Systems Foundation; 2023.","ieee":"M. Safaryan, A. Krumes, and D.-A. Alistarh, “Knowledge distillation performs partial variance reduction,” in <i>36th Conference on Neural Information Processing Systems</i>, New Orleans, LA, United States, 2023, vol. 36.","ista":"Safaryan M, Krumes A, Alistarh D-A. 2023. Knowledge distillation performs partial variance reduction. 36th Conference on Neural Information Processing Systems. NeurIPS: Neural Information Processing Systems, Advances in Neural Information Processing Systems, vol. 36.","chicago":"Safaryan, Mher, Alexandra Krumes, and Dan-Adrian Alistarh. “Knowledge Distillation Performs Partial Variance Reduction.” In <i>36th Conference on Neural Information Processing Systems</i>, Vol. 36. Neural Information Processing Systems Foundation, 2023.","mla":"Safaryan, Mher, et al. “Knowledge Distillation Performs Partial Variance Reduction.” <i>36th Conference on Neural Information Processing Systems</i>, vol. 36, Neural Information Processing Systems Foundation, 2023.","apa":"Safaryan, M., Krumes, A., &#38; Alistarh, D.-A. (2023). Knowledge distillation performs partial variance reduction. In <i>36th Conference on Neural Information Processing Systems</i> (Vol. 36). New Orleans, LA, United States: Neural Information Processing Systems Foundation."},"publisher":"Neural Information Processing Systems Foundation","file":[{"creator":"dernst","file_name":"2023_Neurips_Safaryan.pdf","file_id":"15417","date_created":"2024-05-22T08:08:08Z","checksum":"288c5148a85abf24ad5e22a6b1183655","relation":"main_file","success":1,"content_type":"application/pdf","date_updated":"2024-05-22T08:08:08Z","access_level":"open_access","file_size":672571}],"tmp":{"name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","image":"/images/cc_by.png","short":"CC BY (4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"arxiv":1,"language":[{"iso":"eng"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","ddc":["000"],"publication":"36th Conference on Neural Information Processing Systems","department":[{"_id":"DaAl"}],"acknowledgement":"MS has received funding from the European Union’s Horizon 2020 research and innovation programme\r\nunder the Marie Skłodowska-Curie grant agreement No 101034413.","author":[{"id":"dd546b39-0804-11ed-9c55-ef075c39778d","full_name":"Safaryan, Mher","first_name":"Mher","last_name":"Safaryan"},{"id":"32D78294-F248-11E8-B48F-1D18A9856A87","full_name":"Peste, Elena-Alexandra","last_name":"Peste","first_name":"Elena-Alexandra"},{"orcid":"0000-0003-3650-940X","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","first_name":"Dan-Adrian","last_name":"Alistarh","full_name":"Alistarh, Dan-Adrian"}],"ec_funded":1},{"isi":1,"das_tickbox":"1","title":"Lincheck: A practical framework for testing concurrent data structures on JVM","status":"public","_id":"14260","abstract":[{"text":"This paper presents Lincheck, a new practical and user-friendly framework for testing concurrent algorithms on the Java Virtual Machine (JVM). Lincheck provides a simple and declarative way to write concurrent tests: instead of describing how to perform the test, users specify what to test by declaring all the operations to examine; the framework automatically handles the rest. As a result, tests written with Lincheck are concise and easy to understand. The framework automatically generates a set of concurrent scenarios, examines them using stress-testing or bounded model checking, and verifies that the results of each invocation are correct. Notably, if an error is detected via model checking, Lincheck provides an easy-to-follow trace to reproduce it, significantly simplifying the bug investigation.\r\n\r\nTo the best of our knowledge, Lincheck is the first production-ready tool on the JVM that offers such a simple way of writing concurrent tests, without requiring special skills or expertise. We successfully integrated Lincheck in the development process of several large projects, such as Kotlin Coroutines, and identified new bugs in popular concurrency libraries, such as a race in Java’s standard ConcurrentLinkedDeque and a liveliness bug in Java’s AbstractQueuedSynchronizer framework, which is used in most of the synchronization primitives. We believe that Lincheck can significantly improve the quality and productivity of concurrent algorithms research and development and become the state-of-the-art tool for checking their correctness.","lang":"eng"}],"date_created":"2023-09-03T22:01:16Z","oa":1,"oa_version":"Published Version","scopus_import":"1","date_updated":"2026-07-07T13:39:36Z","conference":{"name":"CAV: Computer Aided Verification","end_date":"2023-07-22","start_date":"2023-07-17","location":"Paris, France"},"year":"2023","page":"156-169","publication_status":"published","type":"conference","external_id":{"isi":["001310786500008"]},"doi":"10.1007/978-3-031-37706-8_8","volume":13964,"has_accepted_license":"1","publication_identifier":{"eissn":["1611-3349"],"issn":["0302-9743"],"isbn":["9783031377051"]},"alternative_title":["LNCS"],"quality_controlled":"1","month":"07","author":[{"full_name":"Koval, Nikita","first_name":"Nikita","last_name":"Koval","id":"2F4DB10C-F248-11E8-B48F-1D18A9856A87"},{"id":"2e711909-896a-11ed-bdf8-eb0f5a2984c6","last_name":"Fedorov","first_name":"Alexander","full_name":"Fedorov, Alexander"},{"first_name":"Maria","last_name":"Sokolova","full_name":"Sokolova, Maria"},{"full_name":"Tsitelov, Dmitry","first_name":"Dmitry","last_name":"Tsitelov"},{"first_name":"Dan-Adrian","last_name":"Alistarh","full_name":"Alistarh, Dan-Adrian","orcid":"0000-0003-3650-940X","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87"}],"related_material":{"record":[{"id":"14995","relation":"research_data","status":"public"}]},"department":[{"_id":"DaAl"},{"_id":"GradSch"}],"language":[{"iso":"eng"}],"publication":"35th International Conference on Computer Aided Verification","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","ddc":["000"],"tmp":{"name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","image":"/images/cc_by.png","short":"CC BY (4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"publisher":"Springer Nature","file":[{"access_level":"open_access","file_size":421408,"content_type":"application/pdf","date_updated":"2023-09-06T08:16:25Z","success":1,"relation":"main_file","checksum":"c346016393123a0a2338ad4d976f61bc","date_created":"2023-09-06T08:16:25Z","file_id":"14275","creator":"dernst","file_name":"2023_LNCS_Koval.pdf"}],"citation":{"ieee":"N. Koval, A. Fedorov, M. Sokolova, D. Tsitelov, and D.-A. Alistarh, “Lincheck: A practical framework for testing concurrent data structures on JVM,” in <i>35th International Conference on Computer Aided Verification</i>, Paris, France, 2023, vol. 13964, pp. 156–169.","chicago":"Koval, Nikita, Alexander Fedorov, Maria Sokolova, Dmitry Tsitelov, and Dan-Adrian Alistarh. “Lincheck: A Practical Framework for Testing Concurrent Data Structures on JVM.” In <i>35th International Conference on Computer Aided Verification</i>, 13964:156–69. Springer Nature, 2023. <a href=\"https://doi.org/10.1007/978-3-031-37706-8_8\">https://doi.org/10.1007/978-3-031-37706-8_8</a>.","ista":"Koval N, Fedorov A, Sokolova M, Tsitelov D, Alistarh D-A. 2023. Lincheck: A practical framework for testing concurrent data structures on JVM. 35th International Conference on Computer Aided Verification. CAV: Computer Aided Verification, LNCS, vol. 13964, 156–169.","ama":"Koval N, Fedorov A, Sokolova M, Tsitelov D, Alistarh D-A. Lincheck: A practical framework for testing concurrent data structures on JVM. In: <i>35th International Conference on Computer Aided Verification</i>. Vol 13964. Springer Nature; 2023:156-169. doi:<a href=\"https://doi.org/10.1007/978-3-031-37706-8_8\">10.1007/978-3-031-37706-8_8</a>","short":"N. Koval, A. Fedorov, M. Sokolova, D. Tsitelov, D.-A. Alistarh, in:, 35th International Conference on Computer Aided Verification, Springer Nature, 2023, pp. 156–169.","apa":"Koval, N., Fedorov, A., Sokolova, M., Tsitelov, D., &#38; Alistarh, D.-A. (2023). Lincheck: A practical framework for testing concurrent data structures on JVM. In <i>35th International Conference on Computer Aided Verification</i> (Vol. 13964, pp. 156–169). Paris, France: Springer Nature. <a href=\"https://doi.org/10.1007/978-3-031-37706-8_8\">https://doi.org/10.1007/978-3-031-37706-8_8</a>","mla":"Koval, Nikita, et al. “Lincheck: A Practical Framework for Testing Concurrent Data Structures on JVM.” <i>35th International Conference on Computer Aided Verification</i>, vol. 13964, Springer Nature, 2023, pp. 156–69, doi:<a href=\"https://doi.org/10.1007/978-3-031-37706-8_8\">10.1007/978-3-031-37706-8_8</a>."},"intvolume":"     13964","file_date_updated":"2023-09-06T08:16:25Z","day":"17","date_published":"2023-07-17T00:00:00Z","article_processing_charge":"Yes (in subscription journal)"},{"department":[{"_id":"DaAl"}],"related_material":{"record":[{"id":"14260","status":"public","relation":"used_in_publication"}]},"oa_version":"Published Version","author":[{"id":"2F4DB10C-F248-11E8-B48F-1D18A9856A87","last_name":"Koval","first_name":"Nikita","full_name":"Koval, Nikita"},{"first_name":"Alexander","last_name":"Fedorov","full_name":"Fedorov, Alexander","id":"2e711909-896a-11ed-bdf8-eb0f5a2984c6"},{"first_name":"Maria","last_name":"Sokolova","full_name":"Sokolova, Maria"},{"full_name":"Tsitelov, Dmitry","first_name":"Dmitry","last_name":"Tsitelov"},{"id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-3650-940X","full_name":"Alistarh, Dan-Adrian","last_name":"Alistarh","first_name":"Dan-Adrian"}],"ddc":["000"],"date_updated":"2026-07-07T13:39:35Z","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","year":"2023","citation":{"apa":"Koval, N., Fedorov, A., Sokolova, M., Tsitelov, D., &#38; Alistarh, D.-A. (2023). Lincheck: A practical framework for testing concurrent data structures on JVM. Zenodo. <a href=\"https://doi.org/10.5281/ZENODO.7877757\">https://doi.org/10.5281/ZENODO.7877757</a>","mla":"Koval, Nikita, et al. <i>Lincheck: A Practical Framework for Testing Concurrent Data Structures on JVM</i>. Zenodo, 2023, doi:<a href=\"https://doi.org/10.5281/ZENODO.7877757\">10.5281/ZENODO.7877757</a>.","short":"N. Koval, A. Fedorov, M. Sokolova, D. Tsitelov, D.-A. Alistarh, (2023).","ista":"Koval N, Fedorov A, Sokolova M, Tsitelov D, Alistarh D-A. 2023. Lincheck: A practical framework for testing concurrent data structures on JVM, Zenodo, <a href=\"https://doi.org/10.5281/ZENODO.7877757\">10.5281/ZENODO.7877757</a>.","ieee":"N. Koval, A. Fedorov, M. Sokolova, D. Tsitelov, and D.-A. Alistarh, “Lincheck: A practical framework for testing concurrent data structures on JVM.” Zenodo, 2023.","chicago":"Koval, Nikita, Alexander Fedorov, Maria Sokolova, Dmitry Tsitelov, and Dan-Adrian Alistarh. “Lincheck: A Practical Framework for Testing Concurrent Data Structures on JVM.” Zenodo, 2023. <a href=\"https://doi.org/10.5281/ZENODO.7877757\">https://doi.org/10.5281/ZENODO.7877757</a>.","ama":"Koval N, Fedorov A, Sokolova M, Tsitelov D, Alistarh D-A. Lincheck: A practical framework for testing concurrent data structures on JVM. 2023. doi:<a href=\"https://doi.org/10.5281/ZENODO.7877757\">10.5281/ZENODO.7877757</a>"},"publisher":"Zenodo","article_processing_charge":"No","day":"28","date_published":"2023-04-28T00:00:00Z","main_file_link":[{"open_access":"1","url":"https://doi.org/10.5281/zenodo.7877757"}],"type":"research_data_reference","doi":"10.5281/ZENODO.7877757","status":"public","_id":"14995","title":"Lincheck: A practical framework for testing concurrent data structures on JVM","oa":1,"date_created":"2024-02-14T15:14:13Z","month":"04","abstract":[{"lang":"eng","text":"Lincheck is a new practical and user-friendly framework for testing concurrent data structures on the Java Virtual Machine (JVM). It provides a simple and declarative way to write concurrent tests. Instead of describing how to perform the test, users specify what to test by declaring all the operations to examine; the framework automatically handles the rest. As a result, tests written with Lincheck are concise and easy to understand. \r\nThe artifact presents a collection of Lincheck tests that discover new bugs in popular libraries and implementations from the concurrency literature -- they are listed in Table 1, Section 3. To evaluate the performance of Lincheck analysis, the collection of tests also includes those which check correct data structures and, thus, always succeed. Similarly to Table 2, Section 3, the experiments demonstrate the reasonable time to perform a test. Finally, Lincheck provides user-friendly output with an easy-to-follow trace to reproduce a detected error, significantly simplifying further investigation."}]},{"project":[{"_id":"9B9290DE-BA93-11EA-9121-9846C619BF3A","name":"Vienna Graduate School on Computational Optimization","grant_number":"W1260-N35"},{"_id":"268A44D6-B435-11E9-9278-68D0E5697425","call_identifier":"H2020","grant_number":"805223","name":"Elastic Coordination for Scalable Machine Learning"}],"year":"2023","publication_status":"published","page":"24364-24373","oa_version":"Preprint","date_updated":"2026-07-27T12:50:03Z","conference":{"location":"Vancouver, BC, Canada","start_date":"2023-06-17","name":"CVPR: Conference on Computer Vision and Pattern Recognition","end_date":"2023-06-24"},"title":"Bias in pruned vision models: In-depth analysis and countermeasures","_id":"14771","status":"public","oa":1,"date_created":"2024-01-10T08:42:40Z","abstract":[{"lang":"eng","text":"Pruning—that is, setting a significant subset of the parameters of a neural network to zero—is one of the most popular methods of model compression. Yet, several recent works have raised the issue that pruning may induce or exacerbate bias in the output of the compressed model. Despite existing evidence for this phenomenon, the relationship between neural network pruning and induced bias is not well-understood. In this work, we systematically investigate and characterize this phenomenon in Convolutional Neural Networks for computer vision. First, we show that it is in fact possible to obtain highly-sparse models, e.g. with less than 10% remaining weights, which do not decrease in accuracy nor substantially increase in bias when compared to dense models. At the same time, we also find that, at higher sparsities, pruned models exhibit higher uncertainty in their outputs, as well as increased correlations, which we directly link to increased bias. We propose easy-to-use criteria which, based only on the uncompressed model, establish whether bias will increase with pruning, and identify the samples most susceptible to biased predictions post-compression. Our code can be found at https://github.com/IST-DASLab/pruned-vision-model-bias."}],"isi":1,"corr_author":"1","citation":{"ieee":"E. B. Iofinova, A. Krumes, and D.-A. Alistarh, “Bias in pruned vision models: In-depth analysis and countermeasures,” in <i>2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition</i>, Vancouver, BC, Canada, 2023, pp. 24364–24373.","ista":"Iofinova EB, Krumes A, Alistarh D-A. 2023. Bias in pruned vision models: In-depth analysis and countermeasures. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. CVPR: Conference on Computer Vision and Pattern Recognition, 24364–24373.","chicago":"Iofinova, Eugenia B, Alexandra Krumes, and Dan-Adrian Alistarh. “Bias in Pruned Vision Models: In-Depth Analysis and Countermeasures.” In <i>2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition</i>, 24364–73. IEEE, 2023. <a href=\"https://doi.org/10.1109/cvpr52729.2023.02334\">https://doi.org/10.1109/cvpr52729.2023.02334</a>.","ama":"Iofinova EB, Krumes A, Alistarh D-A. Bias in pruned vision models: In-depth analysis and countermeasures. In: <i>2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition</i>. IEEE; 2023:24364-24373. doi:<a href=\"https://doi.org/10.1109/cvpr52729.2023.02334\">10.1109/cvpr52729.2023.02334</a>","short":"E.B. Iofinova, A. Krumes, D.-A. Alistarh, in:, 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, IEEE, 2023, pp. 24364–24373.","apa":"Iofinova, E. B., Krumes, A., &#38; Alistarh, D.-A. (2023). Bias in pruned vision models: In-depth analysis and countermeasures. In <i>2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition</i> (pp. 24364–24373). Vancouver, BC, Canada: IEEE. <a href=\"https://doi.org/10.1109/cvpr52729.2023.02334\">https://doi.org/10.1109/cvpr52729.2023.02334</a>","mla":"Iofinova, Eugenia B., et al. “Bias in Pruned Vision Models: In-Depth Analysis and Countermeasures.” <i>2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition</i>, IEEE, 2023, pp. 24364–73, doi:<a href=\"https://doi.org/10.1109/cvpr52729.2023.02334\">10.1109/cvpr52729.2023.02334</a>."},"publisher":"IEEE","article_processing_charge":"No","date_published":"2023-08-22T00:00:00Z","main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2304.12622"}],"day":"22","ec_funded":1,"author":[{"orcid":"0000-0002-7778-3221","id":"f9a17499-f6e0-11ea-865d-fdf9a3f77117","first_name":"Eugenia B","last_name":"Iofinova","full_name":"Iofinova, Eugenia B"},{"full_name":"Peste, Elena-Alexandra","first_name":"Elena-Alexandra","last_name":"Peste","id":"32D78294-F248-11E8-B48F-1D18A9856A87"},{"full_name":"Alistarh, Dan-Adrian","last_name":"Alistarh","first_name":"Dan-Adrian","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-3650-940X"}],"department":[{"_id":"DaAl"},{"_id":"ChLa"}],"related_material":{"record":[{"id":"21854","status":"public","relation":"dissertation_contains"}],"link":[{"relation":"software","url":"https://github.com/IST-DASLab/pruned-vision-model-bias"}]},"acknowledgement":"The authors would like to sincerely thank Sara Hooker for her feedback during the development of this work. EI was supported in part by the FWF DK VGSCO, grant agreement number W1260-N35. AP and DA acknowledge generous ERC support, via Starting Grant 805223 ScaleML.","arxiv":1,"publication":"2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition","language":[{"iso":"eng"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","publication_identifier":{"eisbn":["9798350301298"],"eissn":["2575-7075"]},"month":"08","quality_controlled":"1","external_id":{"isi":["001062531308068"],"arxiv":["2304.12622"]},"type":"conference","doi":"10.1109/cvpr52729.2023.02334"},{"date_updated":"2026-07-29T13:48:39Z","conference":{"start_date":"2023-05-01","location":"Kigali, Rwanda","end_date":"2023-05-05","name":"ICLR: International Conference on Learning Representations"},"scopus_import":"1","oa_version":"Published Version","publication_status":"published","project":[{"call_identifier":"H2020","_id":"268A44D6-B435-11E9-9278-68D0E5697425","name":"Elastic Coordination for Scalable Machine Learning","grant_number":"805223"}],"year":"2023","corr_author":"1","abstract":[{"lang":"eng","text":"Generative Pre-trained Transformer models, known as GPT or OPT, set themselves apart through breakthrough performance across complex language modelling tasks, but also by their extremely high computational and storage costs. Specifically, due to their massive size, even inference for large, highly-accurate GPT models may require multiple performant GPUs, which limits the usability of such models. While there is emerging work on relieving this pressure via model compression, the applicability and performance of existing compression techniques is limited by the scale and complexity of GPT models. In this paper, we address this challenge, and propose OPTQ, a new one-shot weight quantization method based on approximate second-order information, that is both highly-accurate and highly-efficient. Specifically, OPTQ can quantize GPT models with 175 billion parameters in approximately four GPU hours, reducing the bitwidth down to 3 or 4 bits per weight, with negligible accuracy degradation relative to the uncompressed baseline. Our method more than doubles the compression gains relative to previously-proposed one-shot quantization methods, preserving accuracy, allowing us for the first time to execute an 175 billion-parameter model inside a single GPU for generative inference. Moreover, we also show that our method can still provide reasonable accuracy in the extreme quantization regime, in which weights are quantized to 2-bit or even ternary quantization levels. We show experimentally that these improvements can be leveraged for end-to-end inference speedups over FP16, of around 3.25x when using high-end GPUs (NVIDIA A100) and 4.5x when using more cost-effective ones (NVIDIA A6000). The implementation is available at https://github.com/IST-DASLab/gptq."}],"date_created":"2024-08-04T22:01:22Z","oa":1,"title":"OPTQ: Accurate post-training quantization for generative pre-trained transformers","_id":"17378","status":"public","publication":"11th International Conference on Learning Representations ","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","language":[{"iso":"eng"}],"ddc":["000"],"author":[{"id":"09a8f98d-ec99-11ea-ae11-c063a7b7fe5f","full_name":"Frantar, Elias","first_name":"Elias","last_name":"Frantar"},{"first_name":"Saleh","last_name":"Ashkboos","full_name":"Ashkboos, Saleh"},{"first_name":"Torsten","last_name":"Hoefler","full_name":"Hoefler, Torsten"},{"full_name":"Alistarh, Dan-Adrian","last_name":"Alistarh","first_name":"Dan-Adrian","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-3650-940X"}],"ec_funded":1,"acknowledged_ssus":[{"_id":"ScienComp"}],"acknowledgement":"Elias Frantar and Dan Alistarh gratefully acknowledge funding from the European Research Council (ERC) under the European Union’s Horizon 2020 programme (grant agreement No. 805223 ScaleML), as well as experimental support from Eldar Kurtic, and from the IST Austria IT department, in particular Stefano Elefante, Andrei Hornoiu, and Alois Schloegl. The work of Saleh Ashkboos and Torsten Hoefler was supported by the PASC DaCeMI project, received EuroHPC-JU funding under grant MAELSTROM, No. 955513. We thank the Swiss National Supercomputing Center (CSCS) for supporting us with compute infrastructure.","related_material":{"link":[{"url":"https://github.com/IST-DASLab/gptq","relation":"software"}],"record":[{"id":"17485","relation":"dissertation_contains","status":"public"}]},"department":[{"_id":"DaAl"}],"file_date_updated":"2024-08-05T07:52:44Z","date_published":"2023-05-01T00:00:00Z","day":"01","article_processing_charge":"No","file":[{"file_name":"2023_ICLR_Frantar.pdf","creator":"dernst","file_id":"17385","date_created":"2024-08-05T07:52:44Z","relation":"main_file","checksum":"aacbf11dbd8b02a3e0bfd942a33e0593","success":1,"content_type":"application/pdf","date_updated":"2024-08-05T07:52:44Z","access_level":"open_access","file_size":437492}],"publisher":"International Conference on Learning Representations","citation":{"mla":"Frantar, Elias, et al. “OPTQ: Accurate Post-Training Quantization for Generative Pre-Trained Transformers.” <i>11th International Conference on Learning Representations </i>, International Conference on Learning Representations, 2023.","apa":"Frantar, E., Ashkboos, S., Hoefler, T., &#38; Alistarh, D.-A. (2023). OPTQ: Accurate post-training quantization for generative pre-trained transformers. In <i>11th International Conference on Learning Representations </i>. Kigali, Rwanda: International Conference on Learning Representations.","short":"E. Frantar, S. Ashkboos, T. Hoefler, D.-A. Alistarh, in:, 11th International Conference on Learning Representations , International Conference on Learning Representations, 2023.","ama":"Frantar E, Ashkboos S, Hoefler T, Alistarh D-A. OPTQ: Accurate post-training quantization for generative pre-trained transformers. In: <i>11th International Conference on Learning Representations </i>. International Conference on Learning Representations; 2023.","chicago":"Frantar, Elias, Saleh Ashkboos, Torsten Hoefler, and Dan-Adrian Alistarh. “OPTQ: Accurate Post-Training Quantization for Generative Pre-Trained Transformers.” In <i>11th International Conference on Learning Representations </i>. International Conference on Learning Representations, 2023.","ista":"Frantar E, Ashkboos S, Hoefler T, Alistarh D-A. 2023. OPTQ: Accurate post-training quantization for generative pre-trained transformers. 11th International Conference on Learning Representations . ICLR: International Conference on Learning Representations.","ieee":"E. Frantar, S. Ashkboos, T. Hoefler, and D.-A. Alistarh, “OPTQ: Accurate post-training quantization for generative pre-trained transformers,” in <i>11th International Conference on Learning Representations </i>, Kigali, Rwanda, 2023."},"has_accepted_license":"1","type":"conference","quality_controlled":"1","month":"05"}]