[{"acknowledgement":"The authors were partially supported by the 2019 LopezLoreta prize, and they would like to thank (in alphabetical order) Grigorios Chrysos, Simone Maria Giancola, Mahyar\r\nJafari Nodeh, Christoph Lampert, Marco Miani, GuanWen Qiu, and Peter Sukenık for helpful discussions.","page":"4267-4299","publication_status":"published","title":"How spurious features are memorized: Precise analysis for random and NTK features","abstract":[{"lang":"eng","text":"Deep learning models are known to overfit and memorize spurious features in the training dataset. While numerous empirical studies have aimed at understanding this phenomenon, a rigorous theoretical framework to quantify it is still missing. In this paper, we consider spurious features that are uncorrelated with the learning task, and we provide a precise characterization of how they are memorized via two separate terms: (i) the stability of the model with respect to individual training samples, and (ii) the feature alignment between the spurious pattern and the full sample. While the first term is well established in learning theory and it is connected to the generalization error in classical work, the second one is, to the best of our knowledge, novel. Our key technical result gives a precise characterization of the feature alignment for the two prototypical settings of random features (RF) and neural tangent kernel (NTK) regression. We prove that the memorization of spurious features weakens as the generalization capability increases and, through the analysis of the feature alignment, we unveil the role of the model and of its activation function. Numerical experiments show the predictive power of our theory on standard datasets (MNIST, CIFAR-10)."}],"day":"30","author":[{"full_name":"Bombari, Simone","first_name":"Simone","id":"ca726dda-de17-11ea-bc14-f9da834f63aa","last_name":"Bombari"},{"first_name":"Marco","full_name":"Mondelli, Marco","last_name":"Mondelli","id":"27EB676C-8706-11E9-9510-7717E6697425","orcid":"0000-0002-3242-7020"}],"article_processing_charge":"No","OA_type":"green","OA_place":"repository","oa":1,"year":"2024","language":[{"iso":"eng"}],"external_id":{"arxiv":["2305.12100"]},"publication_identifier":{"eissn":["2640-3498"]},"publication":"41st International Conference on Machine Learning","main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2305.12100"}],"project":[{"name":"Prix Lopez-Loretta 2019 - Marco Mondelli","_id":"059876FA-7A3F-11EA-A408-12923DDC885E"}],"scopus_import":"1","date_updated":"2025-04-15T07:50:12Z","date_created":"2025-01-30T07:29:47Z","volume":235,"quality_controlled":"1","_id":"18972","oa_version":"Preprint","month":"07","type":"conference","conference":{"start_date":"2024-07-21","end_date":"2024-07-27","name":"ICML: International Conference on Machine Learning","location":"Vienna, Austria"},"alternative_title":["PMLR"],"arxiv":1,"corr_author":"1","status":"public","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","intvolume":"       235","citation":{"ieee":"S. Bombari and M. Mondelli, “How spurious features are memorized: Precise analysis for random and NTK features,” in <i>41st International Conference on Machine Learning</i>, Vienna, Austria, 2024, vol. 235, pp. 4267–4299.","apa":"Bombari, S., &#38; Mondelli, M. (2024). How spurious features are memorized: Precise analysis for random and NTK features. In <i>41st International Conference on Machine Learning</i> (Vol. 235, pp. 4267–4299). Vienna, Austria: ML Research Press.","ama":"Bombari S, Mondelli M. How spurious features are memorized: Precise analysis for random and NTK features. In: <i>41st International Conference on Machine Learning</i>. Vol 235. ML Research Press; 2024:4267-4299.","chicago":"Bombari, Simone, and Marco Mondelli. “How Spurious Features Are Memorized: Precise Analysis for Random and NTK Features.” In <i>41st International Conference on Machine Learning</i>, 235:4267–99. ML Research Press, 2024.","short":"S. Bombari, M. Mondelli, in:, 41st International Conference on Machine Learning, ML Research Press, 2024, pp. 4267–4299.","mla":"Bombari, Simone, and Marco Mondelli. “How Spurious Features Are Memorized: Precise Analysis for Random and NTK Features.” <i>41st International Conference on Machine Learning</i>, vol. 235, ML Research Press, 2024, pp. 4267–99.","ista":"Bombari S, Mondelli M. 2024. How spurious features are memorized: Precise analysis for random and NTK features. 41st International Conference on Machine Learning. ICML: International Conference on Machine Learning, PMLR, vol. 235, 4267–4299."},"date_published":"2024-07-30T00:00:00Z","publisher":"ML Research Press","department":[{"_id":"MaMo"}]},{"publication":"41st International Conference on Machine Learning","publication_identifier":{"eissn":["2640-3498"]},"external_id":{"arxiv":["2402.02969"]},"language":[{"iso":"eng"}],"year":"2024","OA_place":"repository","oa":1,"OA_type":"green","article_processing_charge":"No","author":[{"full_name":"Bombari, Simone","first_name":"Simone","last_name":"Bombari","id":"ca726dda-de17-11ea-bc14-f9da834f63aa"},{"id":"27EB676C-8706-11E9-9510-7717E6697425","last_name":"Mondelli","orcid":"0000-0002-3242-7020","first_name":"Marco","full_name":"Mondelli, Marco"}],"day":"30","abstract":[{"lang":"eng","text":"Understanding the reasons behind the exceptional success of transformers requires a better analysis of why attention layers are suitable for NLP tasks. In particular, such tasks require predictive models to capture contextual meaning which often depends on one or few words, even if the sentence is long. Our work studies this key property, dubbed word sensitivity (WS), in the prototypical setting of random features. We show that attention layers enjoy high WS, namely, there exists a vector in the space of embeddings that largely perturbs the random attention features map. The argument critically exploits the role of the softmax in the attention layer, highlighting its benefit compared to other activations (e.g., ReLU). In contrast, the WS of standard random features is of order 1/n−−√, n being the number of words in the textual sample, and thus it decays with the length of the context. We then translate these results on the word sensitivity into generalization bounds: due to their low WS, random features provably cannot learn to distinguish between two sentences that differ only in a single word; in contrast, due to their high WS, random attention features have higher generalization capabilities. We validate our theoretical results with experimental evidence over the BERT-Base word embeddings of the imdb review dataset."}],"title":"Towards understanding the word sensitivity of attention layers: A study via random features","publication_status":"published","page":"4300-4328","acknowledgement":"The authors were partially supported by the 2019 LopezLoreta prize, and they would like to thank Mohammad Hossein Amani, Lorenzo Beretta, and Clement Rebuffel for helpful discussions.","department":[{"_id":"MaMo"}],"publisher":"ML Research Press","date_published":"2024-07-30T00:00:00Z","citation":{"ama":"Bombari S, Mondelli M. Towards understanding the word sensitivity of attention layers: A study via random features. In: <i>41st International Conference on Machine Learning</i>. Vol 235. ML Research Press; 2024:4300-4328.","apa":"Bombari, S., &#38; Mondelli, M. (2024). Towards understanding the word sensitivity of attention layers: A study via random features. In <i>41st International Conference on Machine Learning</i> (Vol. 235, pp. 4300–4328). Vienna, Austria: ML Research Press.","ieee":"S. Bombari and M. Mondelli, “Towards understanding the word sensitivity of attention layers: A study via random features,” in <i>41st International Conference on Machine Learning</i>, Vienna, Austria, 2024, vol. 235, pp. 4300–4328.","ista":"Bombari S, Mondelli M. 2024. Towards understanding the word sensitivity of attention layers: A study via random features. 41st International Conference on Machine Learning. ICML: International Conference on Machine Learning, PMLR, vol. 235, 4300–4328.","mla":"Bombari, Simone, and Marco Mondelli. “Towards Understanding the Word Sensitivity of Attention Layers: A Study via Random Features.” <i>41st International Conference on Machine Learning</i>, vol. 235, ML Research Press, 2024, pp. 4300–28.","chicago":"Bombari, Simone, and Marco Mondelli. “Towards Understanding the Word Sensitivity of Attention Layers: A Study via Random Features.” In <i>41st International Conference on Machine Learning</i>, 235:4300–4328. ML Research Press, 2024.","short":"S. Bombari, M. Mondelli, in:, 41st International Conference on Machine Learning, ML Research Press, 2024, pp. 4300–4328."},"intvolume":"       235","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","corr_author":"1","status":"public","arxiv":1,"alternative_title":["PMLR"],"conference":{"end_date":"2024-07-27","name":"ICML: International Conference on Machine Learning","location":"Vienna, Austria","start_date":"2024-07-21"},"type":"conference","month":"07","_id":"18973","oa_version":"Preprint","date_created":"2025-01-30T07:35:49Z","volume":235,"date_updated":"2025-04-15T07:50:12Z","quality_controlled":"1","scopus_import":"1","project":[{"name":"Prix Lopez-Loretta 2019 - Marco Mondelli","_id":"059876FA-7A3F-11EA-A408-12923DDC885E"}],"main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2402.02969"}]},{"day":"29","author":[{"first_name":"Jakub","full_name":"Svoboda, Jakub","last_name":"Svoboda","id":"130759D2-D7DD-11E9-87D2-DE0DE6697425","orcid":"0000-0002-1419-3267"},{"last_name":"Bansal","full_name":"Bansal, Suguman","first_name":"Suguman"},{"first_name":"Krishnendu","full_name":"Chatterjee, Krishnendu","id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87","last_name":"Chatterjee","orcid":"0000-0002-4561-241X"}],"oa":1,"OA_place":"publisher","article_processing_charge":"No","OA_type":"green","page":"47331-47344","publication_status":"published","abstract":[{"lang":"eng","text":"Reinforcement Learning (RL) from temporal logical specifications is a fundamental problem in sequential decision making. One of the basic and core such specification is the reachability specification that requires a target set to be eventually visited. Despite strong empirical results for RL from such specifications, the theoretical guarantees are bleak, including the impossibility of Probably Approximately Correct (PAC) guarantee for reachability specifications. Given the impossibility result, in this work we consider the problem of RL from reachability specifications along with the information of expected conditional distance (ECD). We present (a) lower bound results which establish the necessity of ECD information for PAC guarantees and (b) an algorithm that establishes PAC-guarantees given the ECD information. To the best of our knowledge, this is the first RL from reachability specifications that does not make any assumptions on the underlying environment to learn policies."}],"title":"Reinforcement learning from reachability specifications: PAC guarantees with expected conditional distance","publication":"41st International Conference on Machine Learning","year":"2024","language":[{"iso":"eng"}],"conference":{"start_date":"2024-07-21","end_date":"2024-07-27","location":"Vienna, Austria","name":"ICML: International Conference on Machine Learning"},"type":"conference","main_file_link":[{"url":"https://openreview.net/forum?id=mXUDDL4r1Q","open_access":"1"}],"date_updated":"2025-01-30T07:46:16Z","volume":235,"quality_controlled":"1","date_created":"2025-01-30T07:45:22Z","_id":"18974","oa_version":"Preprint","month":"07","scopus_import":"1","publisher":"ML Research Press","department":[{"_id":"KrCh"}],"corr_author":"1","status":"public","alternative_title":["PMLR"],"date_published":"2024-07-29T00:00:00Z","citation":{"ama":"Svoboda J, Bansal S, Chatterjee K. Reinforcement learning from reachability specifications: PAC guarantees with expected conditional distance. In: <i>41st International Conference on Machine Learning</i>. Vol 235. ML Research Press; 2024:47331-47344.","ieee":"J. Svoboda, S. Bansal, and K. Chatterjee, “Reinforcement learning from reachability specifications: PAC guarantees with expected conditional distance,” in <i>41st International Conference on Machine Learning</i>, Vienna, Austria, 2024, vol. 235, pp. 47331–47344.","apa":"Svoboda, J., Bansal, S., &#38; Chatterjee, K. (2024). Reinforcement learning from reachability specifications: PAC guarantees with expected conditional distance. In <i>41st International Conference on Machine Learning</i> (Vol. 235, pp. 47331–47344). Vienna, Austria: ML Research Press.","ista":"Svoboda J, Bansal S, Chatterjee K. 2024. Reinforcement learning from reachability specifications: PAC guarantees with expected conditional distance. 41st International Conference on Machine Learning. ICML: International Conference on Machine Learning, PMLR, vol. 235, 47331–47344.","short":"J. Svoboda, S. Bansal, K. Chatterjee, in:, 41st International Conference on Machine Learning, ML Research Press, 2024, pp. 47331–47344.","chicago":"Svoboda, Jakub, Suguman Bansal, and Krishnendu Chatterjee. “Reinforcement Learning from Reachability Specifications: PAC Guarantees with Expected Conditional Distance.” In <i>41st International Conference on Machine Learning</i>, 235:47331–44. ML Research Press, 2024.","mla":"Svoboda, Jakub, et al. “Reinforcement Learning from Reachability Specifications: PAC Guarantees with Expected Conditional Distance.” <i>41st International Conference on Machine Learning</i>, vol. 235, ML Research Press, 2024, pp. 47331–44."},"intvolume":"       235","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87"},{"author":[{"last_name":"Modoranu","id":"449f7a18-f128-11eb-9611-9b430c0c6333","first_name":"Ionut-Vlad","full_name":"Modoranu, Ionut-Vlad"},{"first_name":"Aleksei","full_name":"Kalinov, Aleksei","last_name":"Kalinov","id":"44b7120e-eb97-11eb-a6c2-e1557aa81d02","orcid":"0000-0003-2189-3904"},{"first_name":"Eldar","full_name":"Kurtic, Eldar","id":"47beb3a5-07b5-11eb-9b87-b108ec578218","last_name":"Kurtic"},{"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","last_name":"Alistarh","first_name":"Dan-Adrian","full_name":"Alistarh, Dan-Adrian"}],"day":"30","OA_type":"green","article_processing_charge":"No","OA_place":"repository","oa":1,"acknowledgement":"The authors thank Adrian Vladu, Razvan Pascanu, Alexandra Peste, Mher Safaryan for their valuable feedback, the IT department from Institute of Science and Technology Austria for the hardware support and Weights and Biases for the infrastructure to track all our experiments.","publication_status":"published","page":"35910-35933","title":"Error feedback can accurately compress preconditioners","abstract":[{"lang":"eng","text":"Leveraging second-order information about the loss at the scale of deep networks is one of the main lines of approach for improving the performance of current optimizers for deep learning. Yet, existing approaches for accurate full-matrix preconditioning, such as Full-Matrix Adagrad (GGT) or Matrix-Free Approximate Curvature (M-FAC) suffer from massive storage costs when applied even to small-scale models, as they must store a sliding window of gradients, whose memory requirements are multiplicative in the model dimension. In this paper, we address this issue via a novel and efficient error-feedback technique that can be applied to compress preconditioners by up to two orders of magnitude in practice, without loss of convergence. Specifically, our approach compresses the gradient information via sparsification or low-rank compression before it is fed into the preconditioner, feeding the compression error back into future iterations. Extensive experiments on deep neural networks show that this approach can compress full-matrix preconditioners to up to 99% sparsity without accuracy loss, effectively removing the memory overhead of fullmatrix preconditioners such as GGT and M-FAC."}],"acknowledged_ssus":[{"_id":"CampIT"}],"publication_identifier":{"eissn":["2640-3498"]},"publication":"41st International Conference on Machine Learning","year":"2024","external_id":{"arxiv":["2306.06098"]},"language":[{"iso":"eng"}],"type":"conference","conference":{"end_date":"2024-07-27","location":"Vienna, Austria","name":"ICML: International Conference on Machine Learning","start_date":"2024-07-21"},"main_file_link":[{"url":"https://doi.org/10.48550/arXiv.2306.06098","open_access":"1"}],"scopus_import":"1","_id":"18975","month":"07","oa_version":"Preprint","quality_controlled":"1","date_created":"2025-01-30T07:53:22Z","volume":235,"date_updated":"2025-01-30T07:54:16Z","publisher":"ML Research Press","department":[{"_id":"DaAl"}],"arxiv":1,"alternative_title":["PMLR"],"status":"public","corr_author":"1","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","intvolume":"       235","date_published":"2024-07-30T00:00:00Z","citation":{"ama":"Modoranu I-V, Kalinov A, Kurtic E, Frantar E, Alistarh D-A. Error feedback can accurately compress preconditioners. In: <i>41st International Conference on Machine Learning</i>. Vol 235. ML Research Press; 2024:35910-35933.","apa":"Modoranu, I.-V., Kalinov, A., Kurtic, E., Frantar, E., &#38; Alistarh, D.-A. (2024). Error feedback can accurately compress preconditioners. In <i>41st International Conference on Machine Learning</i> (Vol. 235, pp. 35910–35933). Vienna, Austria: ML Research Press.","ieee":"I.-V. Modoranu, A. Kalinov, E. Kurtic, E. Frantar, and D.-A. Alistarh, “Error feedback can accurately compress preconditioners,” in <i>41st International Conference on Machine Learning</i>, Vienna, Austria, 2024, vol. 235, pp. 35910–35933.","ista":"Modoranu I-V, Kalinov A, Kurtic E, Frantar E, Alistarh D-A. 2024. Error feedback can accurately compress preconditioners. 41st International Conference on Machine Learning. ICML: International Conference on Machine Learning, PMLR, vol. 235, 35910–35933.","mla":"Modoranu, Ionut-Vlad, et al. “Error Feedback Can Accurately Compress Preconditioners.” <i>41st International Conference on Machine Learning</i>, vol. 235, ML Research Press, 2024, pp. 35910–33.","short":"I.-V. Modoranu, A. Kalinov, E. Kurtic, E. Frantar, D.-A. Alistarh, in:, 41st International Conference on Machine Learning, ML Research Press, 2024, pp. 35910–35933.","chicago":"Modoranu, Ionut-Vlad, Aleksei Kalinov, Eldar Kurtic, Elias Frantar, and Dan-Adrian Alistarh. “Error Feedback Can Accurately Compress Preconditioners.” In <i>41st International Conference on Machine Learning</i>, 235:35910–33. ML Research Press, 2024."}},{"alternative_title":["PMLR"],"arxiv":1,"ec_funded":1,"corr_author":"1","status":"public","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","intvolume":"       238","citation":{"chicago":"Islamov, Rustem, Mher Safaryan, and Dan-Adrian Alistarh. “AsGrad: A Sharp Unified Analysis of Asynchronous-SGD Algorithms.” In <i>Proceedings of The 27th International Conference on Artificial Intelligence and Statistics</i>, 238:649–57. ML Research Press, 2024.","short":"R. Islamov, M. Safaryan, D.-A. Alistarh, in:, Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, ML Research Press, 2024, pp. 649–657.","mla":"Islamov, Rustem, et al. “AsGrad: A Sharp Unified Analysis of Asynchronous-SGD Algorithms.” <i>Proceedings of The 27th International Conference on Artificial Intelligence and Statistics</i>, vol. 238, ML Research Press, 2024, pp. 649–57.","ista":"Islamov R, Safaryan M, Alistarh D-A. 2024. AsGrad: A sharp unified analysis of asynchronous-SGD algorithms. Proceedings of The 27th International Conference on Artificial Intelligence and Statistics. AISTATS: Conference on Artificial Intelligence and Statistics, PMLR, vol. 238, 649–657.","ieee":"R. Islamov, M. Safaryan, and D.-A. Alistarh, “AsGrad: A sharp unified analysis of asynchronous-SGD algorithms,” in <i>Proceedings of The 27th International Conference on Artificial Intelligence and Statistics</i>, Valencia, Spain, 2024, vol. 238, pp. 649–657.","apa":"Islamov, R., Safaryan, M., &#38; Alistarh, D.-A. (2024). AsGrad: A sharp unified analysis of asynchronous-SGD algorithms. In <i>Proceedings of The 27th International Conference on Artificial Intelligence and Statistics</i> (Vol. 238, pp. 649–657). Valencia, Spain: ML Research Press.","ama":"Islamov R, Safaryan M, Alistarh D-A. AsGrad: A sharp unified analysis of asynchronous-SGD algorithms. In: <i>Proceedings of The 27th International Conference on Artificial Intelligence and Statistics</i>. Vol 238. ML Research Press; 2024:649-657."},"date_published":"2024-05-15T00:00:00Z","publisher":"ML Research Press","department":[{"_id":"DaAl"}],"main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2310.20452"}],"project":[{"grant_number":"101034413","call_identifier":"H2020","name":"IST-BRIDGE: International postdoctoral program","_id":"fc2ed2f7-9c52-11eb-aca3-c01059dda49c"}],"scopus_import":"1","date_updated":"2025-04-14T07:54:52Z","volume":238,"quality_controlled":"1","date_created":"2025-01-30T08:15:49Z","month":"05","_id":"18976","oa_version":"Preprint","type":"conference","conference":{"start_date":"2024-05-02","location":"Valencia, Spain","name":"AISTATS: Conference on Artificial Intelligence and Statistics","end_date":"2024-05-04"},"year":"2024","language":[{"iso":"eng"}],"external_id":{"arxiv":["2310.20452"]},"publication_identifier":{"eissn":["2640-3498"]},"publication":"Proceedings of The 27th International Conference on Artificial Intelligence and Statistics","acknowledgement":"The authors thank all anonymous reviewers for their valuable comments and suggestions on how to improve the manuscript. This work was done when Rustem Islamov was a Master’s student at Institut Polytechnique de Paris (IP Paris) and an intern at Institute of Science and Technology Austria (ISTA). The research of Rustem Islamov was supported by ISTA internship\r\nprogram. Mher Safaryan has received funding from the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No 101034413.","page":"649-657","publication_status":"published","title":"AsGrad: A sharp unified analysis of asynchronous-SGD algorithms","abstract":[{"lang":"eng","text":"We analyze asynchronous-type algorithms for distributed SGD in the heterogeneous setting, where each worker has its own computation and communication speeds, as well as data distribution. In these algorithms, workers compute possibly stale and stochastic gradients associated with their local data at some iteration back in history and then return those gradients to the server without synchronizing with other workers. We present a unified convergence theory for non-convex smooth functions in the heterogeneous regime. The proposed analysis provides convergence for pure asynchronous SGD and its various modifications. Moreover, our theory explains what affects the convergence rate and what can be done to improve the performance of asynchronous algorithms. In particular, we introduce a novel asynchronous method based on worker shuffling. As a by-product of our analysis, we also demonstrate convergence guarantees for gradient-type algorithms such as SGD with random reshuffling and shuffle-once mini-batch SGD. The derived rates match the best-known results for those algorithms, highlighting the tightness of our approach. Finally, our numerical evaluations support theoretical findings and show the good practical performance of our method."}],"day":"15","author":[{"full_name":"Islamov, Rustem","first_name":"Rustem","last_name":"Islamov"},{"first_name":"Mher","full_name":"Safaryan, Mher","id":"dd546b39-0804-11ed-9c55-ef075c39778d","last_name":"Safaryan"},{"first_name":"Dan-Adrian","full_name":"Alistarh, Dan-Adrian","orcid":"0000-0003-3650-940X","last_name":"Alistarh","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87"}],"article_processing_charge":"No","OA_type":"green","oa":1,"OA_place":"repository"},{"year":"2024","language":[{"iso":"eng"}],"external_id":{"arxiv":["2306.03078"]},"publication":"12th International Conference on Learning Representations","publication_status":"published","acknowledgement":"Denis Kuznedelev acknowledges the support from the Russian Ministry of Science and Higher\r\nEducation, grant No. 075-10-2021-068. Ruslan Svirschevski and Vage Egiazarian and Denis\r\nKuznedelev were supported by the grant for research centers in the field of AI provided by the\r\nAnalytical Center for the Government of the Russian Federation (ACRF) in accordance with the\r\nagreement on the provision of subsidies (identifier of the agreement 000000D730321P5Q0002) and the agreement with HSE University No. 70-2021-00139.","abstract":[{"text":"Recent advances in large language model (LLM) pretraining have led to high-quality LLMs with impressive abilities. By compressing such LLMs via quantization to 3-4 bits per parameter, they can fit into memory-limited devices such as laptops and mobile phones, enabling personalized use. Quantizing models to 3-4 bits per parameter can lead to moderate to high accuracy losses, especially for smaller models (1-10B parameters), which are suitable for edge deployment. To address this accuracy issue, we introduce the Sparse-Quantized Representation (SpQR), a new compressed format and quantization technique that enables for the first time \\emph{near-lossless} compression of LLMs across model scales while reaching similar compression levels to previous methods. SpQR works by identifying and isolating \\emph{outlier weights}, which cause particularly large quantization errors, and storing them in higher precision while compressing all other weights to 3-4 bits, and achieves relative accuracy losses of less than \r\n in perplexity for highly-accurate LLaMA and Falcon LLMs. This makes it possible to run a 33B parameter LLM on a single 24 GB consumer GPU without performance degradation at 15% speedup, thus making powerful LLMs available to consumers without any downsides. SpQR comes with efficient algorithms for both encoding weights into its format, as well as decoding them efficiently at runtime. Specifically, we provide an efficient GPU inference algorithm for SpQR, which yields faster inference than 16-bit baselines at similar accuracy while enabling memory compression gains of more than 4x.","lang":"eng"}],"title":"SpQR: A sparse-quantized representation for near-lossless LLM weight compression","day":"15","author":[{"last_name":"Dettmers","full_name":"Dettmers, Tim","first_name":"Tim"},{"last_name":"Svirschevski","full_name":"Svirschevski, Ruslan A.","first_name":"Ruslan A."},{"full_name":"Egiazarian, Vage","first_name":"Vage","last_name":"Egiazarian"},{"first_name":"Denis","full_name":"Kuznedelev, Denis","last_name":"Kuznedelev"},{"id":"09a8f98d-ec99-11ea-ae11-c063a7b7fe5f","last_name":"Frantar","first_name":"Elias","full_name":"Frantar, Elias"},{"last_name":"Ashkboos","first_name":"Saleh","full_name":"Ashkboos, Saleh"},{"last_name":"Borzunov","full_name":"Borzunov, Alexander","first_name":"Alexander"},{"full_name":"Hoefler, Torsten","first_name":"Torsten","last_name":"Hoefler"},{"full_name":"Alistarh, Dan-Adrian","first_name":"Dan-Adrian","orcid":"0000-0003-3650-940X","last_name":"Alistarh","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87"}],"oa":1,"OA_place":"repository","article_processing_charge":"No","OA_type":"green","status":"public","arxiv":1,"citation":{"ama":"Dettmers T, Svirschevski RA, Egiazarian V, et al. SpQR: A sparse-quantized representation for near-lossless LLM weight compression. In: <i>12th International Conference on Learning Representations</i>. OpenReview; 2024.","ieee":"T. Dettmers <i>et al.</i>, “SpQR: A sparse-quantized representation for near-lossless LLM weight compression,” in <i>12th International Conference on Learning Representations</i>, Vienna, Austria, 2024.","apa":"Dettmers, T., Svirschevski, R. A., Egiazarian, V., Kuznedelev, D., Frantar, E., Ashkboos, S., … Alistarh, D.-A. (2024). SpQR: A sparse-quantized representation for near-lossless LLM weight compression. In <i>12th International Conference on Learning Representations</i>. Vienna, Austria: OpenReview.","ista":"Dettmers T, Svirschevski RA, Egiazarian V, Kuznedelev D, Frantar E, Ashkboos S, Borzunov A, Hoefler T, Alistarh D-A. 2024. SpQR: A sparse-quantized representation for near-lossless LLM weight compression. 12th International Conference on Learning Representations. ICLR: International Conference on Learning Representations.","short":"T. Dettmers, R.A. Svirschevski, V. Egiazarian, D. Kuznedelev, E. Frantar, S. Ashkboos, A. Borzunov, T. Hoefler, D.-A. Alistarh, in:, 12th International Conference on Learning Representations, OpenReview, 2024.","chicago":"Dettmers, Tim, Ruslan A. Svirschevski, Vage Egiazarian, Denis Kuznedelev, Elias Frantar, Saleh Ashkboos, Alexander Borzunov, Torsten Hoefler, and Dan-Adrian Alistarh. “SpQR: A Sparse-Quantized Representation for near-Lossless LLM Weight Compression.” In <i>12th International Conference on Learning Representations</i>. OpenReview, 2024.","mla":"Dettmers, Tim, et al. “SpQR: A Sparse-Quantized Representation for near-Lossless LLM Weight Compression.” <i>12th International Conference on Learning Representations</i>, OpenReview, 2024."},"date_published":"2024-05-15T00:00:00Z","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","publisher":"OpenReview","department":[{"_id":"DaAl"}],"main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2306.03078"}],"date_created":"2025-01-30T08:26:59Z","quality_controlled":"1","date_updated":"2025-01-30T08:27:47Z","_id":"18977","month":"05","oa_version":"Preprint","scopus_import":"1","conference":{"start_date":"2024-05-07","end_date":"2024-05-11","location":"Vienna, Austria","name":"ICLR: International Conference on Learning Representations"},"type":"conference"},{"related_material":{"record":[{"relation":"later_version","id":"20323","status":"public"},{"status":"public","id":"18979","relation":"dissertation_contains"}]},"type":"preprint","tmp":{"image":"/images/cc_by.png","short":"CC BY (4.0)","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"doi":"10.48550/arXiv.2209.14993","main_file_link":[{"url":"https://doi.org/10.48550/arXiv.2209.14993","open_access":"1"}],"project":[{"_id":"266A2E9E-B435-11E9-9278-68D0E5697425","name":"Alpha Shape Theory Extended","call_identifier":"H2020","grant_number":"788183"},{"grant_number":"Z00342","call_identifier":"FWF","name":"Mathematics, Computer Science","_id":"268116B8-B435-11E9-9278-68D0E5697425"},{"name":"Persistence and stability of geometric complexes","_id":"2561EBF4-B435-11E9-9278-68D0E5697425","grant_number":"I02979-N35","call_identifier":"FWF"}],"date_created":"2025-01-31T17:03:04Z","date_updated":"2026-04-07T11:47:29Z","_id":"18981","month":"06","oa_version":"Preprint","department":[{"_id":"HeEd"}],"arxiv":1,"ec_funded":1,"status":"public","corr_author":"1","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","date_published":"2024-06-09T00:00:00Z","citation":{"apa":"Brown, A., &#38; Draganov, O. (n.d.). Discrete microlocal Morse theory. <i>arXiv</i>. <a href=\"https://doi.org/10.48550/arXiv.2209.14993\">https://doi.org/10.48550/arXiv.2209.14993</a>","ieee":"A. Brown and O. Draganov, “Discrete microlocal Morse theory,” <i>arXiv</i>. .","ama":"Brown A, Draganov O. Discrete microlocal Morse theory. <i>arXiv</i>. doi:<a href=\"https://doi.org/10.48550/arXiv.2209.14993\">10.48550/arXiv.2209.14993</a>","mla":"Brown, Adam, and Ondrej Draganov. “Discrete Microlocal Morse Theory.” <i>ArXiv</i>, doi:<a href=\"https://doi.org/10.48550/arXiv.2209.14993\">10.48550/arXiv.2209.14993</a>.","chicago":"Brown, Adam, and Ondrej Draganov. “Discrete Microlocal Morse Theory.” <i>ArXiv</i>, n.d. <a href=\"https://doi.org/10.48550/arXiv.2209.14993\">https://doi.org/10.48550/arXiv.2209.14993</a>.","short":"A. Brown, O. Draganov, ArXiv (n.d.).","ista":"Brown A, Draganov O. Discrete microlocal Morse theory. arXiv, <a href=\"https://doi.org/10.48550/arXiv.2209.14993\">10.48550/arXiv.2209.14993</a>."},"day":"09","author":[{"last_name":"Brown","first_name":"Adam","full_name":"Brown, Adam"},{"id":"2B23F01E-F248-11E8-B48F-1D18A9856A87","last_name":"Draganov","orcid":"0000-0003-0464-3823","full_name":"Draganov, Ondrej","first_name":"Ondrej"}],"article_processing_charge":"No","oa":1,"OA_place":"repository","acknowledgement":"This project has received funding from the European Research Council (ERC) under the European\r\nUnion’s Horizon 2020 research and innovation programme, grant no. 788183, from the Wittgenstein Prize,\r\nAustrian Science Fund (FWF), grant no. Z 342-N31, and from the DFG Collaborative Research Center TRR\r\n109, ‘Discretization in Geometry and Dynamics’, Austrian Science Fund (FWF), grant no. I 02979-N35.","publication_status":"draft","title":"Discrete microlocal Morse theory","abstract":[{"lang":"eng","text":"We establish several results combining discrete Morse theory and microlocal sheaf theory in the setting of finite posets and simplicial complexes. Our primary tool is a computationally tractable description of the bounded derived category of sheaves on a poset with the Alexandrov topology. We prove that each bounded complex of sheaves on a finite poset admits a unique (up to isomorphism of complexes) minimal injective resolution, and we provide algorithms for computing minimal injective resolution of an injective complex, as well as several useful functors between derived categories of sheaves. For the constant sheaf on a simplicial complex, we give asymptotically tight bounds on the complexity of computing the minimal injective resolution using those algorithms. Our main result is a novel definition of the discrete microsupport of a bounded complex of sheaves on a finite poset. We detail several foundational properties of the discrete microsupport, as well as a microlocal generalization of the discrete homological Morse theorem and Morse inequalities."}],"publication":"arXiv","year":"2024","language":[{"iso":"eng"}],"external_id":{"arxiv":["2209.14993"]}},{"publisher":"Neural Information Processing Systems Foundation","department":[{"_id":"FrLo"}],"file_date_updated":"2025-02-04T13:09:08Z","alternative_title":["Advances in Neural Information Processing Systems"],"arxiv":1,"status":"public","intvolume":"        37","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","date_published":"2024-09-25T00:00:00Z","citation":{"ista":"Chen T, Bello K, Locatello F, Aragam B, Ravikumar PK. 2024. Identifying general mechanism shifts in linear causal representations. 38th Conference on Neural Information Processing Systems. NeurIPS: Neural Information Processing Systems, Advances in Neural Information Processing Systems, vol. 37.","chicago":"Chen, Tianyu, Kevin Bello, Francesco Locatello, Bryon Aragam, and Pradeep Kumar Ravikumar. “Identifying General Mechanism Shifts in Linear Causal Representations.” In <i>38th Conference on Neural Information Processing Systems</i>, Vol. 37. Neural Information Processing Systems Foundation, 2024.","short":"T. Chen, K. Bello, F. Locatello, B. Aragam, P.K. Ravikumar, in:, 38th Conference on Neural Information Processing Systems, Neural Information Processing Systems Foundation, 2024.","mla":"Chen, Tianyu, et al. “Identifying General Mechanism Shifts in Linear Causal Representations.” <i>38th Conference on Neural Information Processing Systems</i>, vol. 37, Neural Information Processing Systems Foundation, 2024.","ama":"Chen T, Bello K, Locatello F, Aragam B, Ravikumar PK. Identifying general mechanism shifts in linear causal representations. In: <i>38th Conference on Neural Information Processing Systems</i>. Vol 37. Neural Information Processing Systems Foundation; 2024.","ieee":"T. Chen, K. Bello, F. Locatello, B. Aragam, and P. K. Ravikumar, “Identifying general mechanism shifts in linear causal representations,” in <i>38th Conference on Neural Information Processing Systems</i>, Vancouver, Canada, 2024, vol. 37.","apa":"Chen, T., Bello, K., Locatello, F., Aragam, B., &#38; Ravikumar, P. K. (2024). Identifying general mechanism shifts in linear causal representations. In <i>38th Conference on Neural Information Processing Systems</i> (Vol. 37). Vancouver, Canada: Neural Information Processing Systems Foundation."},"type":"conference","has_accepted_license":"1","tmp":{"image":"/images/cc_by.png","short":"CC BY (4.0)","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"conference":{"end_date":"2024-12-16","name":"NeurIPS: Neural Information Processing Systems","location":"Vancouver, Canada","start_date":"2024-12-16"},"scopus_import":"1","quality_controlled":"1","date_created":"2025-02-04T13:09:34Z","volume":37,"date_updated":"2025-07-07T13:23:49Z","_id":"18996","oa_version":"Published Version","month":"09","file":[{"content_type":"application/pdf","access_level":"open_access","success":1,"creator":"dernst","file_size":5659119,"date_created":"2025-02-04T13:09:08Z","date_updated":"2025-02-04T13:09:08Z","checksum":"75c3091e70bd2916cd94afbf40a0c425","file_name":"2024_NeurIPS_Chen.pdf","file_id":"18997","relation":"main_file"}],"publication_identifier":{"eissn":["1049-5258"]},"publication":"38th Conference on Neural Information Processing Systems","year":"2024","language":[{"iso":"eng"}],"external_id":{"arxiv":["2410.24059"]},"day":"25","ddc":["000"],"author":[{"last_name":"Chen","full_name":"Chen, Tianyu","first_name":"Tianyu"},{"last_name":"Bello","full_name":"Bello, Kevin","first_name":"Kevin"},{"orcid":"0000-0002-4850-0683","last_name":"Locatello","id":"26cfd52f-2483-11ee-8040-88983bcc06d4","full_name":"Locatello, Francesco","first_name":"Francesco"},{"first_name":"Bryon","full_name":"Aragam, Bryon","last_name":"Aragam"},{"last_name":"Ravikumar","full_name":"Ravikumar, Pradeep Kumar","first_name":"Pradeep Kumar"}],"article_processing_charge":"No","OA_type":"green","OA_place":"repository","oa":1,"publication_status":"published","title":"Identifying general mechanism shifts in linear causal representations","abstract":[{"text":"We consider the linear causal representation learning setting where we observe a linear mixing of d unknown latent factors, which follow a linear structural causal model. Recent work has shown that it is possible to recover the latent factors as well as the underlying structural causal model over them, up to permutation and scaling, provided that we have at least d environments, each of which corresponds to perfect interventions on a single latent node (factor). After this powerful result, a key open problem faced by the community has been to relax these conditions: allow for coarser than perfect single-node interventions, and allow for fewer than d of them, since the number of latent factors d could be very large. In this work, we consider precisely such a setting, where we allow a smaller than d number of environments, and also allow for very coarse interventions that can very coarsely \\textit{change the entire causal graph over the latent factors}. On the flip side, we relax what we wish to extract to simply the \\textit{list of nodes that have shifted between one or more environments}. We provide a surprising identifiability result that it is indeed possible, under some very mild standard assumptions, to identify the set of shifted nodes. Our identifiability proof moreover is a constructive one: we explicitly provide necessary and sufficient conditions for a node to be a shifted node, and show that we can check these conditions given observed data. Our algorithm lends itself very naturally to the sample setting where instead of just interventional distributions, we are provided datasets of samples from each of these distributions. We corroborate our results on both synthetic experiments as well as an interesting psychometric dataset. The code can be found at https://github.com/TianyuCodings/iLCS.","lang":"eng"}]},{"language":[{"iso":"eng"}],"external_id":{"arxiv":["2404.00500"]},"year":"2024","publication":"Findings of the Association for Computational Linguistics: EMNLP 2024","file":[{"file_name":"2024_EMNLP_Draganov.pdf","relation":"main_file","file_id":"19016","checksum":"f4416a5962194f0181ab0dc7f9ef93c0","creator":"dernst","success":1,"file_size":1312638,"date_updated":"2025-02-10T08:20:34Z","date_created":"2025-02-10T08:20:34Z","access_level":"open_access","content_type":"application/pdf"}],"title":"The shape of word embeddings: Quantifying non-isometry with topological data analysis","abstract":[{"lang":"eng","text":"Word embeddings represent language vocabularies as clouds of d-dimensional points. We investigate how information is conveyed by the general shape of these clouds, instead of representing the semantic meaning of each token. Specifically, we use the notion of persistent homology from topological data analysis (TDA) to measure the distances between language pairs from the shape of their unlabeled embeddings. These distances quantify the degree of non-isometry of the embeddings. To distinguish whether these differences are random training errors or capture real information about the languages, we use the computed distance matrices to construct language phylogenetic trees over 81 Indo-European languages. Careful evaluation shows that our reconstructed trees exhibit strong and statistically-significant similarities to the reference."}],"page":"12080-12099","publication_status":"published","article_processing_charge":"No","OA_type":"gold","OA_place":"publisher","oa":1,"day":"01","ddc":["500"],"author":[{"orcid":"0000-0003-0464-3823","last_name":"Draganov","id":"2B23F01E-F248-11E8-B48F-1D18A9856A87","full_name":"Draganov, Ondrej","first_name":"Ondrej"},{"last_name":"Skiena","first_name":"Steven","full_name":"Skiena, Steven"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","date_published":"2024-11-01T00:00:00Z","citation":{"chicago":"Draganov, Ondrej, and Steven Skiena. “The Shape of Word Embeddings: Quantifying Non-Isometry with Topological Data Analysis.” In <i>Findings of the Association for Computational Linguistics: EMNLP 2024</i>, 12080–99. Association for Computational Linguistics, 2024. <a href=\"https://doi.org/10.18653/v1/2024.findings-emnlp.705\">https://doi.org/10.18653/v1/2024.findings-emnlp.705</a>.","short":"O. Draganov, S. Skiena, in:, Findings of the Association for Computational Linguistics: EMNLP 2024, Association for Computational Linguistics, 2024, pp. 12080–12099.","mla":"Draganov, Ondrej, and Steven Skiena. “The Shape of Word Embeddings: Quantifying Non-Isometry with Topological Data Analysis.” <i>Findings of the Association for Computational Linguistics: EMNLP 2024</i>, Association for Computational Linguistics, 2024, pp. 12080–99, doi:<a href=\"https://doi.org/10.18653/v1/2024.findings-emnlp.705\">10.18653/v1/2024.findings-emnlp.705</a>.","ista":"Draganov O, Skiena S. 2024. The shape of word embeddings: Quantifying non-isometry with topological data analysis. Findings of the Association for Computational Linguistics: EMNLP 2024. EMNLP: Conference on Empirical Methods in Natural Language Processing, 12080–12099.","ieee":"O. Draganov and S. Skiena, “The shape of word embeddings: Quantifying non-isometry with topological data analysis,” in <i>Findings of the Association for Computational Linguistics: EMNLP 2024</i>, Miami, FL, United States, 2024, pp. 12080–12099.","apa":"Draganov, O., &#38; Skiena, S. (2024). The shape of word embeddings: Quantifying non-isometry with topological data analysis. In <i>Findings of the Association for Computational Linguistics: EMNLP 2024</i> (pp. 12080–12099). Miami, FL, United States: Association for Computational Linguistics. <a href=\"https://doi.org/10.18653/v1/2024.findings-emnlp.705\">https://doi.org/10.18653/v1/2024.findings-emnlp.705</a>","ama":"Draganov O, Skiena S. The shape of word embeddings: Quantifying non-isometry with topological data analysis. In: <i>Findings of the Association for Computational Linguistics: EMNLP 2024</i>. Association for Computational Linguistics; 2024:12080-12099. doi:<a href=\"https://doi.org/10.18653/v1/2024.findings-emnlp.705\">10.18653/v1/2024.findings-emnlp.705</a>"},"arxiv":1,"corr_author":"1","status":"public","department":[{"_id":"GradSch"},{"_id":"HeEd"}],"file_date_updated":"2025-02-10T08:20:34Z","publisher":"Association for Computational Linguistics","scopus_import":"1","date_created":"2025-02-04T16:19:28Z","quality_controlled":"1","date_updated":"2025-02-10T08:21:37Z","_id":"18998","month":"11","oa_version":"Published Version","doi":"10.18653/v1/2024.findings-emnlp.705","type":"conference","tmp":{"image":"/images/cc_by.png","short":"CC BY (4.0)","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"has_accepted_license":"1","conference":{"start_date":"2024-11-12","end_date":"2024-11-16","name":"EMNLP: Conference on Empirical Methods in Natural Language Processing","location":"Miami, FL, United States"}},{"language":[{"iso":"eng"}],"external_id":{"arxiv":["2406.04102"]},"year":"2024","publication":"arXiv","abstract":[{"lang":"eng","text":"Exploring the shape of point configurations has been a key driver in the evolution of TDA (short for topological data analysis) since its infancy. This survey illustrates the recent efforts to broaden these ideas to model spatial interactions among multiple configurations, each distinguished by a color. It describes advances in this area and prepares the ground for further exploration by mentioning unresolved questions and promising research avenues while focusing on the overlap with discrete geometry."}],"title":"Chromatic topological data analysis","publication_status":"submitted","OA_place":"repository","oa":1,"article_processing_charge":"No","OA_type":"green","ddc":["510"],"day":"06","author":[{"first_name":"Sebastiano","full_name":"Cultrera di Montesano, Sebastiano","id":"34D2A09C-F248-11E8-B48F-1D18A9856A87","last_name":"Cultrera di Montesano","orcid":"0000-0001-6249-0832"},{"first_name":"Ondrej","full_name":"Draganov, Ondrej","last_name":"Draganov","id":"2B23F01E-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0003-0464-3823"},{"id":"3FB178DA-F248-11E8-B48F-1D18A9856A87","last_name":"Edelsbrunner","orcid":"0000-0002-9823-6833","full_name":"Edelsbrunner, Herbert","first_name":"Herbert"},{"last_name":"Saghafian","id":"f86f7148-b140-11ec-9577-95435b8df824","full_name":"Saghafian, Morteza","first_name":"Morteza"}],"citation":{"apa":"Cultrera di Montesano, S., Draganov, O., Edelsbrunner, H., &#38; Saghafian, M. (n.d.). Chromatic topological data analysis. <i>arXiv</i>. <a href=\"https://doi.org/10.48550/ARXIV.2406.04102\">https://doi.org/10.48550/ARXIV.2406.04102</a>","ieee":"S. Cultrera di Montesano, O. Draganov, H. Edelsbrunner, and M. Saghafian, “Chromatic topological data analysis,” <i>arXiv</i>. .","ama":"Cultrera di Montesano S, Draganov O, Edelsbrunner H, Saghafian M. Chromatic topological data analysis. <i>arXiv</i>. doi:<a href=\"https://doi.org/10.48550/ARXIV.2406.04102\">10.48550/ARXIV.2406.04102</a>","mla":"Cultrera di Montesano, Sebastiano, et al. “Chromatic Topological Data Analysis.” <i>ArXiv</i>, 2406.04102, doi:<a href=\"https://doi.org/10.48550/ARXIV.2406.04102\">10.48550/ARXIV.2406.04102</a>.","short":"S. Cultrera di Montesano, O. Draganov, H. Edelsbrunner, M. Saghafian, ArXiv (n.d.).","chicago":"Cultrera di Montesano, Sebastiano, Ondrej Draganov, Herbert Edelsbrunner, and Morteza Saghafian. “Chromatic Topological Data Analysis.” <i>ArXiv</i>, n.d. <a href=\"https://doi.org/10.48550/ARXIV.2406.04102\">https://doi.org/10.48550/ARXIV.2406.04102</a>.","ista":"Cultrera di Montesano S, Draganov O, Edelsbrunner H, Saghafian M. Chromatic topological data analysis. arXiv, 2406.04102."},"date_published":"2024-06-06T00:00:00Z","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","corr_author":"1","status":"public","arxiv":1,"article_number":"2406.04102","department":[{"_id":"GradSch"},{"_id":"HeEd"}],"date_created":"2025-02-04T16:21:21Z","date_updated":"2025-02-10T08:14:27Z","month":"06","_id":"18999","oa_version":"Preprint","main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2406.04102"}],"doi":"10.48550/ARXIV.2406.04102","has_accepted_license":"1","tmp":{"image":"/images/cc_by.png","short":"CC BY (4.0)","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"type":"preprint"},{"publication_status":"published","acknowledgement":"We thank Niklas Boers for recommending the SpeedyWeather simulator and Valentino Maiorca\r\nfor guidance on Fourier transformation for SST data. We are also grateful to Shimeng Huang and Riccardo Cadei for their feedback on the treatment effect estimation experiment and to Jiale Chen and Adeel Pervez for their assistance with the solver implementation. Finally, we appreciate the anonymous reviewers for their insightful suggestions, which helped improve the manuscript. ","abstract":[{"text":"Causal representation learning promises to extend causal models to hidden causal\r\nvariables from raw entangled measurements. However, most progress has focused\r\non proving identifiability results in different settings, and we are not aware of any\r\nsuccessful real-world application. At the same time, the field of dynamical systems\r\nbenefited from deep learning and scaled to countless applications but does not allow\r\nparameter identification. In this paper, we draw a clear connection between the two\r\nand their key assumptions, allowing us to apply identifiable methods developed\r\nin causal representation learning to dynamical systems. At the same time, we can\r\nleverage scalable differentiable solvers developed for differential equations to build\r\nmodels that are both identifiable and practical. Overall, we learn explicitly controllable models that isolate the trajectory-specific parameters for further downstream\r\ntasks such as out-of-distribution classification or treatment effect estimation. We\r\nexperiment with a wind simulator with partially known factors of variation. We\r\nalso apply the resulting model to real-world climate data and successfully answer\r\ndownstream causal questions in line with existing literature on climate change.\r\nCode is available at https://github.com/CausalLearningAI/crl-dynamical-systems.","lang":"eng"}],"title":"Marrying causal representation learning with dynamical systems for science","author":[{"id":"d3e02e50-48a8-11ee-8f62-c108061797fa","last_name":"Yao","full_name":"Yao, Dingling","first_name":"Dingling"},{"first_name":"Caroline J","full_name":"Muller, Caroline J","id":"f978ccb0-3f7f-11eb-b193-b0e2bd13182b","last_name":"Muller","orcid":"0000-0001-5836-5350"},{"first_name":"Francesco","full_name":"Locatello, Francesco","orcid":"0000-0002-4850-0683","id":"26cfd52f-2483-11ee-8040-88983bcc06d4","last_name":"Locatello"}],"day":"01","ddc":["000","550"],"oa":1,"OA_place":"publisher","OA_type":"gold","article_processing_charge":"No","year":"2024","external_id":{"arxiv":["2405.13888"]},"language":[{"iso":"eng"}],"file":[{"checksum":"fe8832367e7143876f178244385d859e","file_name":"2024_NeurIPS_Yao.pdf","file_id":"19006","relation":"main_file","content_type":"application/pdf","access_level":"open_access","success":1,"creator":"dernst","date_created":"2025-02-05T07:44:58Z","file_size":2595855,"date_updated":"2025-02-05T07:44:58Z"}],"publication":"38th Conference on Neural Information Processing Systems","_id":"19005","month":"12","oa_version":"Published Version","quality_controlled":"1","date_updated":"2025-07-10T11:51:32Z","volume":37,"date_created":"2025-02-05T07:49:00Z","scopus_import":"1","related_material":{"link":[{"relation":"software","url":"https://github.com/CausalLearningAI/crl-dynamical-systems"}]},"conference":{"end_date":"2024-12-16","name":"NeurIPS: Neural Information Processing Systems","location":"Vancouver, Canada","start_date":"2024-12-16"},"tmp":{"image":"/images/cc_by.png","short":"CC BY (4.0)","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"has_accepted_license":"1","type":"conference","corr_author":"1","status":"public","arxiv":1,"alternative_title":["Advances in Neural Information Processing Systems"],"date_published":"2024-12-01T00:00:00Z","citation":{"apa":"Yao, D., Muller, C. J., &#38; Locatello, F. (2024). Marrying causal representation learning with dynamical systems for science. In <i>38th Conference on Neural Information Processing Systems</i> (Vol. 37). Vancouver, Canada: Neural Information Processing Systems Foundation.","ieee":"D. Yao, C. J. Muller, and F. Locatello, “Marrying causal representation learning with dynamical systems for science,” in <i>38th Conference on Neural Information Processing Systems</i>, Vancouver, Canada, 2024, vol. 37.","ama":"Yao D, Muller CJ, Locatello F. Marrying causal representation learning with dynamical systems for science. In: <i>38th Conference on Neural Information Processing Systems</i>. Vol 37. Neural Information Processing Systems Foundation; 2024.","mla":"Yao, Dingling, et al. “Marrying Causal Representation Learning with Dynamical Systems for Science.” <i>38th Conference on Neural Information Processing Systems</i>, vol. 37, Neural Information Processing Systems Foundation, 2024.","short":"D. Yao, C.J. Muller, F. Locatello, in:, 38th Conference on Neural Information Processing Systems, Neural Information Processing Systems Foundation, 2024.","chicago":"Yao, Dingling, Caroline J Muller, and Francesco Locatello. “Marrying Causal Representation Learning with Dynamical Systems for Science.” In <i>38th Conference on Neural Information Processing Systems</i>, Vol. 37. Neural Information Processing Systems Foundation, 2024.","ista":"Yao D, Muller CJ, Locatello F. 2024. Marrying causal representation learning with dynamical systems for science. 38th Conference on Neural Information Processing Systems. NeurIPS: Neural Information Processing Systems, Advances in Neural Information Processing Systems, vol. 37."},"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","intvolume":"        37","publisher":"Neural Information Processing Systems Foundation","file_date_updated":"2025-02-05T07:44:58Z","department":[{"_id":"CaMu"},{"_id":"FrLo"}]},{"month":"12","_id":"19007","oa_version":"Published Version","quality_controlled":"1","date_created":"2025-02-05T08:36:22Z","volume":37,"date_updated":"2025-05-14T11:29:10Z","scopus_import":"1","conference":{"location":"Vancouver, Canada","name":"NeurIPS: Neural Information Processing Systems","end_date":"2024-12-16","start_date":"2024-12-16"},"has_accepted_license":"1","type":"conference","status":"public","arxiv":1,"alternative_title":["Advances in Neural Information Processing Systems"],"citation":{"ama":"Kori A, Locatello F, Santhirasekaram A, Toni F, Glocker B, De Sousa Ribeiro F. Identifiable object-centric representation learning via probabilistic slot attention. In: <i>38th Conference on Neural Information Processing Systems</i>. Vol 37. Neural Information Processing Systems Foundation; 2024.","ieee":"A. Kori, F. Locatello, A. Santhirasekaram, F. Toni, B. Glocker, and F. De Sousa Ribeiro, “Identifiable object-centric representation learning via probabilistic slot attention,” in <i>38th Conference on Neural Information Processing Systems</i>, Vancouver, Canada, 2024, vol. 37.","apa":"Kori, A., Locatello, F., Santhirasekaram, A., Toni, F., Glocker, B., &#38; De Sousa Ribeiro, F. (2024). Identifiable object-centric representation learning via probabilistic slot attention. In <i>38th Conference on Neural Information Processing Systems</i> (Vol. 37). Vancouver, Canada: Neural Information Processing Systems Foundation.","ista":"Kori A, Locatello F, Santhirasekaram A, Toni F, Glocker B, De Sousa Ribeiro F. 2024. Identifiable object-centric representation learning via probabilistic slot attention. 38th Conference on Neural Information Processing Systems. NeurIPS: Neural Information Processing Systems, Advances in Neural Information Processing Systems, vol. 37.","short":"A. Kori, F. Locatello, A. Santhirasekaram, F. Toni, B. Glocker, F. De Sousa Ribeiro, in:, 38th Conference on Neural Information Processing Systems, Neural Information Processing Systems Foundation, 2024.","chicago":"Kori, Avinash, Francesco Locatello, Ainkaran Santhirasekaram, Francesca Toni, Ben Glocker, and Fabio De Sousa Ribeiro. “Identifiable Object-Centric Representation Learning via Probabilistic Slot Attention.” In <i>38th Conference on Neural Information Processing Systems</i>, Vol. 37. Neural Information Processing Systems Foundation, 2024.","mla":"Kori, Avinash, et al. “Identifiable Object-Centric Representation Learning via Probabilistic Slot Attention.” <i>38th Conference on Neural Information Processing Systems</i>, vol. 37, Neural Information Processing Systems Foundation, 2024."},"date_published":"2024-12-01T00:00:00Z","intvolume":"        37","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","publisher":"Neural Information Processing Systems Foundation","file_date_updated":"2025-02-05T08:34:25Z","department":[{"_id":"FrLo"}],"publication_status":"published","acknowledgement":"A. Kori is supported by UKRI (grant number EP/S023356/1), as part of the UKRI Centre for Doctoral Training in Safe and Trusted AI. B. Glocker and F.D.S. Ribeiro acknowledge the support of the UKRI AI programme, and the Engineering and Physical Sciences Research Council, for CHAI - EPSRC Causality in Healthcare AI Hub (grant number EP/Y028856/1).","abstract":[{"text":"Learning modular object-centric representations is crucial for systematic generalization. Existing methods show promising object-binding capabilities empirically,\r\nbut theoretical identifiability guarantees remain relatively underdeveloped. Understanding when object-centric representations can theoretically be identified is\r\ncrucial for scaling slot-based methods to high-dimensional images with correctness\r\nguarantees. To that end, we propose a probabilistic slot-attention algorithm that\r\nimposes an aggregate mixture prior over object-centric slot representations, thereby\r\nproviding slot identifiability guarantees without supervision, up to an equivalence\r\nrelation. We provide empirical verification of our theoretical identifiability result\r\nusing both simple 2-dimensional data and high-resolution imaging datasets.\r\n","lang":"eng"}],"title":"Identifiable object-centric representation learning via probabilistic slot attention","author":[{"full_name":"Kori, Avinash","first_name":"Avinash","last_name":"Kori"},{"full_name":"Locatello, Francesco","first_name":"Francesco","id":"26cfd52f-2483-11ee-8040-88983bcc06d4","last_name":"Locatello","orcid":"0000-0002-4850-0683"},{"full_name":"Santhirasekaram, Ainkaran","first_name":"Ainkaran","last_name":"Santhirasekaram"},{"last_name":"Toni","first_name":"Francesca","full_name":"Toni, Francesca"},{"full_name":"Glocker, Ben","first_name":"Ben","last_name":"Glocker"},{"first_name":"Fabio","full_name":"De Sousa Ribeiro, Fabio","last_name":"De Sousa Ribeiro"}],"day":"01","ddc":["000"],"oa":1,"OA_place":"publisher","OA_type":"hybrid","article_processing_charge":"No","year":"2024","external_id":{"arxiv":["2406.07141"]},"language":[{"iso":"eng"}],"file":[{"file_name":"2024_NeurIPS_Kori.pdf","relation":"main_file","file_id":"19008","checksum":"d27b3c7102adc28e798fe41001f0b919","creator":"dernst","success":1,"file_size":6943800,"date_updated":"2025-02-05T08:34:25Z","date_created":"2025-02-05T08:34:25Z","access_level":"open_access","content_type":"application/pdf"}],"publication":"38th Conference on Neural Information Processing Systems"},{"title":"Terminal singularities of the moduli space of curves on low degree hypersurfaces and the circle method","month":"12","_id":"19013","oa_version":"Preprint","abstract":[{"text":"We study the singularities of the moduli space of degree e maps from smooth genus g curves to an arbitrary smooth hypersurface of low degree. For e large compared to g, we show that these moduli spaces have at worst terminal singularities. Our main approach is to study the jet schemes of these moduli spaces by developing a suitable form of the circle method.","lang":"eng"}],"date_updated":"2025-04-15T08:05:40Z","date_created":"2025-02-07T12:04:11Z","doi":"10.48550/arXiv.2412.14923","main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2412.14923"}],"publication_status":"draft","type":"preprint","article_processing_charge":"No","OA_place":"repository","tmp":{"image":"/images/cc_by.png","short":"CC BY (4.0)","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"oa":1,"author":[{"last_name":"Glas","id":"d6423cba-dc74-11ea-a0a7-ee61689ff5fb","first_name":"Jakob","full_name":"Glas, Jakob"},{"full_name":"Hase-Liu, Matthew ","first_name":"Matthew ","last_name":"Hase-Liu"}],"day":"19","related_material":{"record":[{"status":"public","relation":"earlier_version","id":"18295"}]},"external_id":{"arxiv":["2412.14923"]},"user_id":"8b945eb4-e2f2-11eb-945a-df72226e66a9","language":[{"iso":"eng"}],"date_published":"2024-12-19T00:00:00Z","citation":{"ama":"Glas J, Hase-Liu M. Terminal singularities of the moduli space of curves on low degree hypersurfaces and the circle method. <i>arXiv</i>. doi:<a href=\"https://doi.org/10.48550/arXiv.2412.14923\">10.48550/arXiv.2412.14923</a>","apa":"Glas, J., &#38; Hase-Liu, M. (n.d.). Terminal singularities of the moduli space of curves on low degree hypersurfaces and the circle method. <i>arXiv</i>. <a href=\"https://doi.org/10.48550/arXiv.2412.14923\">https://doi.org/10.48550/arXiv.2412.14923</a>","ieee":"J. Glas and M. Hase-Liu, “Terminal singularities of the moduli space of curves on low degree hypersurfaces and the circle method,” <i>arXiv</i>. .","ista":"Glas J, Hase-Liu M. Terminal singularities of the moduli space of curves on low degree hypersurfaces and the circle method. arXiv, <a href=\"https://doi.org/10.48550/arXiv.2412.14923\">10.48550/arXiv.2412.14923</a>.","mla":"Glas, Jakob, and Matthew Hase-Liu. “Terminal Singularities of the Moduli Space of Curves on Low Degree Hypersurfaces and the Circle Method.” <i>ArXiv</i>, doi:<a href=\"https://doi.org/10.48550/arXiv.2412.14923\">10.48550/arXiv.2412.14923</a>.","short":"J. Glas, M. Hase-Liu, ArXiv (n.d.).","chicago":"Glas, Jakob, and Matthew  Hase-Liu. “Terminal Singularities of the Moduli Space of Curves on Low Degree Hypersurfaces and the Circle Method.” <i>ArXiv</i>, n.d. <a href=\"https://doi.org/10.48550/arXiv.2412.14923\">https://doi.org/10.48550/arXiv.2412.14923</a>."},"year":"2024","arxiv":1,"corr_author":"1","status":"public","publication":"arXiv","department":[{"_id":"TiBr"}]},{"oa_version":"Published Version","_id":"19051","month":"07","date_updated":"2025-09-09T12:16:45Z","quality_controlled":"1","date_created":"2025-02-18T07:15:50Z","volume":2024,"scopus_import":"1","doi":"10.1093/imrn/rnae066","tmp":{"image":"/images/cc_by.png","short":"CC BY (4.0)","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"has_accepted_license":"1","type":"journal_article","related_material":{"record":[{"status":"public","id":"254","relation":"earlier_version"}]},"citation":{"apa":"Browning, T. D. (2024). The polynomial sieve and equal sums of like polynomials. <i>International Mathematics Research Notices</i>. Oxford University Press. <a href=\"https://doi.org/10.1093/imrn/rnae066\">https://doi.org/10.1093/imrn/rnae066</a>","ieee":"T. D. Browning, “The polynomial sieve and equal sums of like polynomials,” <i>International Mathematics Research Notices</i>, vol. 2024, no. 13. Oxford University Press, pp. 10165–10168, 2024.","ama":"Browning TD. The polynomial sieve and equal sums of like polynomials. <i>International Mathematics Research Notices</i>. 2024;2024(13):10165-10168. doi:<a href=\"https://doi.org/10.1093/imrn/rnae066\">10.1093/imrn/rnae066</a>","mla":"Browning, Timothy D. “The Polynomial Sieve and Equal Sums of like Polynomials.” <i>International Mathematics Research Notices</i>, vol. 2024, no. 13, Oxford University Press, 2024, pp. 10165–68, doi:<a href=\"https://doi.org/10.1093/imrn/rnae066\">10.1093/imrn/rnae066</a>.","short":"T.D. Browning, International Mathematics Research Notices 2024 (2024) 10165–10168.","chicago":"Browning, Timothy D. “The Polynomial Sieve and Equal Sums of like Polynomials.” <i>International Mathematics Research Notices</i>. Oxford University Press, 2024. <a href=\"https://doi.org/10.1093/imrn/rnae066\">https://doi.org/10.1093/imrn/rnae066</a>.","ista":"Browning TD. 2024. The polynomial sieve and equal sums of like polynomials. International Mathematics Research Notices. 2024(13), 10165–10168."},"date_published":"2024-07-01T00:00:00Z","user_id":"317138e5-6ab7-11ef-aa6d-ffef3953e345","intvolume":"      2024","status":"public","corr_author":"1","file_date_updated":"2025-02-18T07:56:36Z","department":[{"_id":"TiBr"}],"publisher":"Oxford University Press","abstract":[{"lang":"eng","text":"This paper corrects an error in an earlier work of the author."}],"title":"The polynomial sieve and equal sums of like polynomials","publication_status":"published","page":"10165-10168","oa":1,"OA_place":"publisher","OA_type":"hybrid","article_processing_charge":"Yes (via OA deal)","issue":"13","author":[{"id":"35827D50-F248-11E8-B48F-1D18A9856A87","last_name":"Browning","orcid":"0000-0002-8314-0177","full_name":"Browning, Timothy D","first_name":"Timothy D"}],"day":"01","ddc":["510"],"external_id":{"isi":["001196957300001"]},"language":[{"iso":"eng"}],"year":"2024","article_type":"original","publication":"International Mathematics Research Notices","publication_identifier":{"eissn":["1687-0247"],"issn":["1073-7928"]},"file":[{"file_name":"2024_IMRN_Browning.pdf","relation":"main_file","file_id":"19052","checksum":"b625b8adf018d2a97591813c1fc17b96","creator":"dernst","success":1,"date_updated":"2025-02-18T07:56:36Z","file_size":205750,"date_created":"2025-02-18T07:56:36Z","access_level":"open_access","content_type":"application/pdf"}],"isi":1},{"oa":1,"OA_place":"repository","article_processing_charge":"No","OA_type":"green","day":"01","ddc":["000"],"author":[{"last_name":"Zverev","id":"05162b19-1340-11ed-8f02-fa94e0e8c3bc","first_name":"Egor","full_name":"Zverev, Egor"},{"last_name":"Abdelnabi","first_name":"Sahar","full_name":"Abdelnabi, Sahar"},{"id":"06000900-6068-11ef-8d61-c2472ef2e752","last_name":"Tabesh","orcid":"0009-0003-4119-6281","full_name":"Tabesh, Soroush","first_name":"Soroush"},{"first_name":"Mario","full_name":"Fritz, Mario","last_name":"Fritz"},{"last_name":"Lampert","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0001-8622-7887","full_name":"Lampert, Christoph","first_name":"Christoph"}],"abstract":[{"lang":"eng","text":"Instruction-tuned Large Language Models (LLMs) show impressive results in numerous practical applications, but they lack essential safety features that are common in other areas of computer science, particularly an explicit separation of instructions and data. This makes them vulnerable to manipulations such as indirect prompt injections and generally unsuitable for safety-critical tasks. Surprisingly, there is currently no established definition or benchmark to quantify this phenomenon. In this work, we close this gap by introducing a formal measure for instruction-data separation and an empirical variant that is calculable from a model's outputs. We also present a new dataset, SEP, that allows estimating the measure for real-world models. Our results on various LLMs show that the problem of instruction-data separation is real: all models fail to achieve high separation, and canonical mitigation techniques, such as prompt engineering and fine-tuning, either fail to substantially improve separation or reduce model utility. The source code and SEP dataset are openly accessible at https://github.com/egozverev/Shold-It-Be-Executed-Or-Processed.\r\n"}],"title":"Can LLMs separate instructions from data? And what do we even mean by that?","publication_status":"published","acknowledgement":"The authors would like to sincerely thank Juan Rocamonde for valuable feedback to our manuscript. We acknowledge the support from the Scientific Service Units (SSU) of ISTA through resources provided by Scientific Computing (SciComp). We thank Dan Alistarh for providing us with computational resources. This work was partially funded by the German Federal Ministry of Education and Research (BMBF) under the grant AIgenCY (16KIS2012) and ELSA – European Lighthouse on Secure and Safe AI funded by the European Union under grant agreement No. 101070617. Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or European Commission. Neither the European Union nor the European Commission can be held responsible for them.","publication":"arXiv","acknowledged_ssus":[{"_id":"ScienComp"}],"file":[{"success":1,"creator":"ezverev","date_created":"2025-02-20T10:11:45Z","file_size":530972,"date_updated":"2025-02-20T10:11:45Z","access_level":"open_access","content_type":"application/pdf","file_name":"2403.06833v3.pdf","relation":"main_file","file_id":"19064","checksum":"35eb43968684b87be59144603ef10af0"}],"language":[{"iso":"eng"}],"external_id":{"arxiv":["2403.06833"]},"year":"2024","has_accepted_license":"1","tmp":{"image":"/images/cc_by_sa.png","name":"Creative Commons Attribution-ShareAlike 4.0 International Public License (CC BY-SA 4.0)","legal_code_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","short":"CC BY-SA (4.0)"},"type":"preprint","related_material":{"link":[{"relation":"software","url":" https://github.com/egozverev/Shold-It-Be-Executed-Or-Processed"}]},"date_updated":"2025-02-24T12:52:23Z","date_created":"2025-02-20T10:13:42Z","month":"03","_id":"19063","oa_version":"Preprint","main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2403.06833"}],"doi":"10.48550/arXiv.2403.06833","department":[{"_id":"GradSch"},{"_id":"ChLa"}],"license":"https://creativecommons.org/licenses/by-sa/4.0/","file_date_updated":"2025-02-20T10:11:45Z","citation":{"ama":"Zverev E, Abdelnabi S, Tabesh S, Fritz M, Lampert C. Can LLMs separate instructions from data? And what do we even mean by that? <i>arXiv</i>. 2024. doi:<a href=\"https://doi.org/10.48550/arXiv.2403.06833\">10.48550/arXiv.2403.06833</a>","ieee":"E. Zverev, S. Abdelnabi, S. Tabesh, M. Fritz, and C. Lampert, “Can LLMs separate instructions from data? And what do we even mean by that?,” <i>arXiv</i>. 2024.","apa":"Zverev, E., Abdelnabi, S., Tabesh, S., Fritz, M., &#38; Lampert, C. (2024). Can LLMs separate instructions from data? And what do we even mean by that? <i>arXiv</i>. <a href=\"https://doi.org/10.48550/arXiv.2403.06833\">https://doi.org/10.48550/arXiv.2403.06833</a>","ista":"Zverev E, Abdelnabi S, Tabesh S, Fritz M, Lampert C. 2024. Can LLMs separate instructions from data? And what do we even mean by that? arXiv, 2403.06833.","chicago":"Zverev, Egor, Sahar Abdelnabi, Soroush Tabesh, Mario Fritz, and Christoph Lampert. “Can LLMs Separate Instructions from Data? And What Do We Even Mean by That?” <i>ArXiv</i>, 2024. <a href=\"https://doi.org/10.48550/arXiv.2403.06833\">https://doi.org/10.48550/arXiv.2403.06833</a>.","short":"E. Zverev, S. Abdelnabi, S. Tabesh, M. Fritz, C. Lampert, ArXiv (2024).","mla":"Zverev, Egor, et al. “Can LLMs Separate Instructions from Data? And What Do We Even Mean by That?” <i>ArXiv</i>, 2403.06833, 2024, doi:<a href=\"https://doi.org/10.48550/arXiv.2403.06833\">10.48550/arXiv.2403.06833</a>."},"date_published":"2024-03-01T00:00:00Z","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","corr_author":"1","status":"public","arxiv":1,"article_number":"2403.06833"},{"year":"2024","status":"public","corr_author":"1","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","date_published":"2024-02-21T00:00:00Z","citation":{"ista":"Hwong Y-L, Muller CJ. 2024. Data - The unreasonable efficiency of total rain evaporation removal in triggering convective self-aggregation, Zenodo, <a href=\"https://doi.org/10.5281/ZENODO.10687169\">10.5281/ZENODO.10687169</a>.","mla":"Hwong, Yi-Ling, and Caroline J. Muller. <i>Data - The Unreasonable Efficiency of Total Rain Evaporation Removal in Triggering Convective Self-Aggregation</i>. Zenodo, 2024, doi:<a href=\"https://doi.org/10.5281/ZENODO.10687169\">10.5281/ZENODO.10687169</a>.","chicago":"Hwong, Yi-Ling, and Caroline J Muller. “Data - The Unreasonable Efficiency of Total Rain Evaporation Removal in Triggering Convective Self-Aggregation.” Zenodo, 2024. <a href=\"https://doi.org/10.5281/ZENODO.10687169\">https://doi.org/10.5281/ZENODO.10687169</a>.","short":"Y.-L. Hwong, C.J. Muller, (2024).","ama":"Hwong Y-L, Muller CJ. Data - The unreasonable efficiency of total rain evaporation removal in triggering convective self-aggregation. 2024. doi:<a href=\"https://doi.org/10.5281/ZENODO.10687169\">10.5281/ZENODO.10687169</a>","apa":"Hwong, Y.-L., &#38; Muller, C. J. (2024). Data - The unreasonable efficiency of total rain evaporation removal in triggering convective self-aggregation. Zenodo. <a href=\"https://doi.org/10.5281/ZENODO.10687169\">https://doi.org/10.5281/ZENODO.10687169</a>","ieee":"Y.-L. Hwong and C. J. Muller, “Data - The unreasonable efficiency of total rain evaporation removal in triggering convective self-aggregation.” Zenodo, 2024."},"publisher":"Zenodo","department":[{"_id":"CaMu"}],"doi":"10.5281/ZENODO.10687169","main_file_link":[{"url":"https://doi.org/10.5281/zenodo.8369509","open_access":"1"}],"title":"Data - The unreasonable efficiency of total rain evaporation removal in triggering convective self-aggregation","date_updated":"2025-09-04T13:16:39Z","abstract":[{"text":"This repository contains the data, scripts, SAM codes and files required to reproduce the results of the manuscript \"The Unreasonable Efficiency of Total Rain Evaporation Removal in Triggering Convective Self-Aggregation\" submitted to the Geophysical Research Letters (GRL).\r\n\r\nBrief description of project: This project aims to examine the impact of rain evaporation removal or reduction in the planetary boundary layer (PBL) on convective self aggregation (CSA). Non-rotating radiative-convective equilibrium (RCE) simulations were conducted with the System for Atmospheric Modeling (SAM) cloud resolving model. Rain evaporation in the lowest 1 km was progressively reduced and the effect on CSA was investigated. The physical processes underlying this type of aggregation (referred to in the manuscript as no-evaporation CSA, or NE-CSA) were analyzed and described. \r\nThe default SAM code base (version 6.10.8) can be downloaded from here: http://rossby.msrc.sunysb.edu/~marat/SAM.html","lang":"eng"}],"date_created":"2025-03-07T08:39:40Z","_id":"19307","month":"02","oa_version":"Published Version","ddc":["550"],"day":"21","related_material":{"record":[{"id":"15186","relation":"used_in_publication","status":"public"}]},"author":[{"first_name":"Yi-Ling","full_name":"Hwong, Yi-Ling","id":"1217aa61-4dd1-11ec-9ac3-f2ba3f17ee22","last_name":"Hwong","orcid":"0000-0001-9281-3479"},{"orcid":"0000-0001-5836-5350","last_name":"Muller","id":"f978ccb0-3f7f-11eb-b193-b0e2bd13182b","full_name":"Muller, Caroline J","first_name":"Caroline J"}],"article_processing_charge":"No","OA_type":"green","type":"research_data_reference","tmp":{"image":"/images/cc_by.png","short":"CC BY (4.0)","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"OA_place":"repository","oa":1,"has_accepted_license":"1"},{"oa":1,"OA_place":"publisher","OA_type":"diamond","article_processing_charge":"No","author":[{"last_name":"Verwimp","full_name":"Verwimp, Eli","first_name":"Eli"},{"last_name":"Aljundi","first_name":"Rahaf","full_name":"Aljundi, Rahaf"},{"full_name":"Ben-David, Shai","first_name":"Shai","last_name":"Ben-David"},{"full_name":"Bethge, Matthias","first_name":"Matthias","last_name":"Bethge"},{"last_name":"Cossu","full_name":"Cossu, Andrea","first_name":"Andrea"},{"last_name":"Gepperth","full_name":"Gepperth, Alexander","first_name":"Alexander"},{"last_name":"Hayes","full_name":"Hayes, Tyler L.","first_name":"Tyler L."},{"full_name":"Hüllermeier, Eyke","first_name":"Eyke","last_name":"Hüllermeier"},{"last_name":"Kanan","first_name":"Christopher","full_name":"Kanan, Christopher"},{"last_name":"Kudithipudi","first_name":"Dhireesha","full_name":"Kudithipudi, Dhireesha"},{"full_name":"Lampert, Christoph","first_name":"Christoph","orcid":"0000-0001-8622-7887","id":"40C20FD2-F248-11E8-B48F-1D18A9856A87","last_name":"Lampert"},{"first_name":"Martin","full_name":"Mundt, Martin","last_name":"Mundt"},{"last_name":"Pascanu","full_name":"Pascanu, Razvan","first_name":"Razvan"},{"last_name":"Popescu","first_name":"Adrian","full_name":"Popescu, Adrian"},{"last_name":"Tolias","first_name":"Andreas S.","full_name":"Tolias, Andreas S."},{"first_name":"Joost","full_name":"Van De Weijer, Joost","last_name":"Van De Weijer"},{"first_name":"Bing","full_name":"Liu, Bing","last_name":"Liu"},{"first_name":"Vincenzo","full_name":"Lomonaco, Vincenzo","last_name":"Lomonaco"},{"last_name":"Tuytelaars","full_name":"Tuytelaars, Tinne","first_name":"Tinne"},{"last_name":"Van De Ven","first_name":"Gido M.","full_name":"Van De Ven, Gido M."}],"ddc":["000"],"day":"12","abstract":[{"text":"Continual learning is a subfield of machine learning, which aims to allow machine learning models to continuously learn on new data, by accumulating knowledge without forgetting what was learned in the past. In this work, we take a step back, and ask: \"Why should one care about continual learning in the first place?\". We set the stage by examining recent continual learning papers published at four major machine learning conferences, and show that memory-constrained settings dominate the field. Then, we discuss five open problems in machine learning, and even though they might seem unrelated to continual learning at first sight, we show that continual learning will inevitably be part of their solution. These problems are model editing, personalization and specialization, on-device learning, faster (re-)training and reinforcement learning. Finally, by comparing the desiderata from these unsolved problems and the current assumptions in continual learning, we highlight and discuss four future directions for continual learning research. We hope that this work offers an interesting perspective on the future of continual learning, while displaying its potential value and the paths we have to pursue in order to make it successful. This work is the result of the many discussions the authors had at the Dagstuhl seminar on Deep Continual Learning, in March 2023.","lang":"eng"}],"title":"Continual learning: Applications and the road forward","publication_status":"published","publication":"Transactions on Machine Learning Research","publication_identifier":{"eissn":["2835-8856"]},"file":[{"access_level":"open_access","content_type":"application/pdf","creator":"dernst","success":1,"date_created":"2025-03-20T09:02:18Z","file_size":1367966,"date_updated":"2025-03-20T09:02:18Z","checksum":"0714e12f7423cd098976ed9974561155","file_name":"2024_TMLR_Verwimp.pdf","relation":"main_file","file_id":"19426"}],"external_id":{"arxiv":["2311.11908"]},"language":[{"iso":"eng"}],"year":"2024","article_type":"original","tmp":{"image":"/images/cc_by.png","short":"CC BY (4.0)","name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode"},"has_accepted_license":"1","type":"journal_article","month":"04","_id":"19408","oa_version":"Published Version","quality_controlled":"1","date_updated":"2025-03-20T09:21:02Z","volume":2024,"date_created":"2025-03-16T23:01:25Z","scopus_import":"1","file_date_updated":"2025-03-20T09:02:18Z","department":[{"_id":"ChLa"}],"publisher":"Transactions on Machine Learning Research","date_published":"2024-04-12T00:00:00Z","citation":{"short":"E. Verwimp, R. Aljundi, S. Ben-David, M. Bethge, A. Cossu, A. Gepperth, T.L. Hayes, E. Hüllermeier, C. Kanan, D. Kudithipudi, C. Lampert, M. Mundt, R. Pascanu, A. Popescu, A.S. Tolias, J. Van De Weijer, B. Liu, V. Lomonaco, T. Tuytelaars, G.M. Van De Ven, Transactions on Machine Learning Research 2024 (2024).","chicago":"Verwimp, Eli, Rahaf Aljundi, Shai Ben-David, Matthias Bethge, Andrea Cossu, Alexander Gepperth, Tyler L. Hayes, et al. “Continual Learning: Applications and the Road Forward.” <i>Transactions on Machine Learning Research</i>. Transactions on Machine Learning Research, 2024.","mla":"Verwimp, Eli, et al. “Continual Learning: Applications and the Road Forward.” <i>Transactions on Machine Learning Research</i>, vol. 2024, Transactions on Machine Learning Research, 2024.","ista":"Verwimp E, Aljundi R, Ben-David S, Bethge M, Cossu A, Gepperth A, Hayes TL, Hüllermeier E, Kanan C, Kudithipudi D, Lampert C, Mundt M, Pascanu R, Popescu A, Tolias AS, Van De Weijer J, Liu B, Lomonaco V, Tuytelaars T, Van De Ven GM. 2024. Continual learning: Applications and the road forward. Transactions on Machine Learning Research. 2024.","ieee":"E. Verwimp <i>et al.</i>, “Continual learning: Applications and the road forward,” <i>Transactions on Machine Learning Research</i>, vol. 2024. Transactions on Machine Learning Research, 2024.","apa":"Verwimp, E., Aljundi, R., Ben-David, S., Bethge, M., Cossu, A., Gepperth, A., … Van De Ven, G. M. (2024). Continual learning: Applications and the road forward. <i>Transactions on Machine Learning Research</i>. Transactions on Machine Learning Research.","ama":"Verwimp E, Aljundi R, Ben-David S, et al. Continual learning: Applications and the road forward. <i>Transactions on Machine Learning Research</i>. 2024;2024."},"intvolume":"      2024","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","status":"public","arxiv":1,"alternative_title":["TMLR"]},{"oa_version":"Preprint","_id":"19425","month":"08","date_updated":"2026-04-07T11:48:53Z","date_created":"2025-03-20T07:48:23Z","project":[{"grant_number":"801770","call_identifier":"H2020","name":"Angulon: physics and applications of a new quasiparticle","_id":"2688CF98-B435-11E9-9278-68D0E5697425"}],"main_file_link":[{"url":"https://doi.org/10.48550/arXiv.2408.16848","open_access":"1"}],"doi":"10.48550/arXiv.2408.16848","type":"preprint","related_material":{"record":[{"id":"19393","relation":"dissertation_contains","status":"public"}]},"date_published":"2024-08-29T00:00:00Z","citation":{"ista":"Karle V, Lemeshko M, Bouhon A, Slager R-J, Ünal FN. Anomalous multi-gap topological phases in periodically driven quantum  rotors. arXiv, 2408.16848.","chicago":"Karle, Volker, Mikhail Lemeshko, Adrien Bouhon, Robert-Jan Slager, and F. Nur Ünal. “Anomalous Multi-Gap Topological Phases in Periodically Driven Quantum  Rotors.” <i>ArXiv</i>, n.d. <a href=\"https://doi.org/10.48550/arXiv.2408.16848\">https://doi.org/10.48550/arXiv.2408.16848</a>.","short":"V. Karle, M. Lemeshko, A. Bouhon, R.-J. Slager, F.N. Ünal, ArXiv (n.d.).","mla":"Karle, Volker, et al. “Anomalous Multi-Gap Topological Phases in Periodically Driven Quantum  Rotors.” <i>ArXiv</i>, 2408.16848, doi:<a href=\"https://doi.org/10.48550/arXiv.2408.16848\">10.48550/arXiv.2408.16848</a>.","ama":"Karle V, Lemeshko M, Bouhon A, Slager R-J, Ünal FN. Anomalous multi-gap topological phases in periodically driven quantum  rotors. <i>arXiv</i>. doi:<a href=\"https://doi.org/10.48550/arXiv.2408.16848\">10.48550/arXiv.2408.16848</a>","ieee":"V. Karle, M. Lemeshko, A. Bouhon, R.-J. Slager, and F. N. Ünal, “Anomalous multi-gap topological phases in periodically driven quantum  rotors,” <i>arXiv</i>. .","apa":"Karle, V., Lemeshko, M., Bouhon, A., Slager, R.-J., &#38; Ünal, F. N. (n.d.). Anomalous multi-gap topological phases in periodically driven quantum  rotors. <i>arXiv</i>. <a href=\"https://doi.org/10.48550/arXiv.2408.16848\">https://doi.org/10.48550/arXiv.2408.16848</a>"},"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","status":"public","ec_funded":1,"arxiv":1,"article_number":"2408.16848","department":[{"_id":"MiLe"}],"abstract":[{"lang":"eng","text":"We demonstrate that periodically driven quantum rotors provide a promising and broadly applicable platform to implement multi-gap topological phases, where groups of bands can acquire topological invariants due to non-Abelian braiding of band degeneracies. By adiabatically varying the periodic kicks to the rotor we find nodal-line braiding, which causes sign flips of topological charges of band nodes and can prevent them from annihilating, indicated by non-zero values of the %non-Abelian patch Euler class. In particular, we report\r\non the emergence of an anomalous Dirac string phase arising in the strongly driven regime, a truly out-of-equilibrium phase of the quantum rotor. This phase emanates from braiding processes involving all (quasienergy) gaps and manifests itself with edge states at zero angular momentum. Our results reveal direct applications in state-of-the-art experiments of quantum rotors, such as linear molecules driven by periodic far-off-resonant laser pulses or artificial\r\nquantum rotors in optical lattices, whose extensive versatility offers precise modification and observation of novel non-Abelian topological properties. "}],"title":"Anomalous multi-gap topological phases in periodically driven quantum  rotors","publication_status":"draft","acknowledgement":"We thank G. M. Koutentakis, S. Wimberger, J. G. E. Harris, T. Enss and A. Ghazaryan for fruitful discussions. M.L. acknowledges support by the European Research Council (ERC) Starting Grant No. 801770 (ANGULON). R.-J. S. acknowledges funding from a EPSRC ERC underwrite grant EP/X025829/1, a EPSRC New Investigator Award grant EP/W00187X/1, as well as Trinity College, Cambridge. F.N.U. acknowledges support from the Marie ¨Sk lodowska-Curie programme of the European Commission [Grant No. 893915], Simons Investigator Award\r\n[Grant No. 511029] and Trinity College Cambridge.","oa":1,"OA_place":"repository","OA_type":"green","article_processing_charge":"No","author":[{"full_name":"Karle, Volker","first_name":"Volker","last_name":"Karle","id":"D7C012AE-D7ED-11E9-95E8-1EC5E5697425","orcid":"0000-0002-6963-0129"},{"first_name":"Mikhail","full_name":"Lemeshko, Mikhail","orcid":"0000-0002-6990-7802","id":"37CB05FA-F248-11E8-B48F-1D18A9856A87","last_name":"Lemeshko"},{"last_name":"Bouhon","first_name":"Adrien","full_name":"Bouhon, Adrien"},{"first_name":"Robert-Jan","full_name":"Slager, Robert-Jan","last_name":"Slager"},{"last_name":"Ünal","full_name":"Ünal, F. Nur","first_name":"F. Nur"}],"day":"29","external_id":{"arxiv":["2408.16848"]},"language":[{"iso":"eng"}],"year":"2024","publication":"arXiv"},{"type":"journal_article","volume":2,"date_created":"2025-03-23T23:01:28Z","quality_controlled":"1","date_updated":"2025-03-25T08:28:39Z","oa_version":"None","_id":"19446","month":"10","scopus_import":"1","doi":"10.1038/s44220-024-00316-z","department":[{"_id":"GaNo"}],"publisher":"Springer Nature","date_published":"2024-10-01T00:00:00Z","citation":{"ama":"Nees F, Renner P, Holz NE, et al. Large-scale population data enrichment in mental health research. <i>Nature Mental Health</i>. 2024;2(10):1124-1127. doi:<a href=\"https://doi.org/10.1038/s44220-024-00316-z\">10.1038/s44220-024-00316-z</a>","apa":"Nees, F., Renner, P., Holz, N. E., Polemiti, E., Siehl, S., Hese, S., … Ogoh, G. (2024). Large-scale population data enrichment in mental health research. <i>Nature Mental Health</i>. Springer Nature. <a href=\"https://doi.org/10.1038/s44220-024-00316-z\">https://doi.org/10.1038/s44220-024-00316-z</a>","ieee":"F. Nees <i>et al.</i>, “Large-scale population data enrichment in mental health research,” <i>Nature Mental Health</i>, vol. 2, no. 10. Springer Nature, pp. 1124–1127, 2024.","ista":"Nees F, Renner P, Holz NE, Polemiti E, Siehl S, Hese S, Schepanski K, Schumann G, Walter H, Heinz A, Ralser M, Twardziok S, Vaidya N, Bernas A, Serin E, Jentsch M, Hitchen E, Kebir H, Lett TA, Roy JC, Eils R, Taron UH, Schütz T, Banks J, Banaschewski T, Jansone K, Christmann N, Meyer-Lindenberg A, Tost H, Holz N, Schwarz E, Stringaris A, Neidhart M, Seefried B, Aden R, Andreassen OA, Westlye LT, Van Der Meer D, Fernandez S, Kjelkenes R, Ask H, Rapp M, Tschorn M, Böttger SJ, Marquand A, Novarino G, Marr L, Slater M, Viapiana GF, Orosa FE, Gallego J, Pastor A, Forstner AJ, Hoffmann P, Nöthen MM, Claus I, Miller A, Mathey CM, Heilmann-Heimbach S, Sommer P, Patraskaki M, Wilbertz J, Schmitt K, Jirsa V, Petkoski S, Pitel S, Otten L, Athanasiadis AP, Pearmund C, Spanlang B, Alvarez E, Sanchez M, Giner A, Jia T, Gong Y, Xia Y, Chang X, Calhoun V, Liu J, Schwalber A, Thompson P, Clinton N, Desrivières S, Young AH, Stahl B, Ogoh G. 2024. Large-scale population data enrichment in mental health research. Nature Mental Health. 2(10), 1124–1127.","mla":"Nees, Frauke, et al. “Large-Scale Population Data Enrichment in Mental Health Research.” <i>Nature Mental Health</i>, vol. 2, no. 10, Springer Nature, 2024, pp. 1124–27, doi:<a href=\"https://doi.org/10.1038/s44220-024-00316-z\">10.1038/s44220-024-00316-z</a>.","chicago":"Nees, Frauke, Paul Renner, Nathalie E. Holz, Elli Polemiti, Sebastian Siehl, Sören Hese, Kerstin Schepanski, et al. “Large-Scale Population Data Enrichment in Mental Health Research.” <i>Nature Mental Health</i>. Springer Nature, 2024. <a href=\"https://doi.org/10.1038/s44220-024-00316-z\">https://doi.org/10.1038/s44220-024-00316-z</a>.","short":"F. Nees, P. Renner, N.E. Holz, E. Polemiti, S. Siehl, S. Hese, K. Schepanski, G. Schumann, H. Walter, A. Heinz, M. Ralser, S. Twardziok, N. Vaidya, A. Bernas, E. Serin, M. Jentsch, E. Hitchen, H. Kebir, T.A. Lett, J.C. Roy, R. Eils, U.H. Taron, T. Schütz, J. Banks, T. Banaschewski, K. Jansone, N. Christmann, A. Meyer-Lindenberg, H. Tost, N. Holz, E. Schwarz, A. Stringaris, M. Neidhart, B. Seefried, R. Aden, O.A. Andreassen, L.T. Westlye, D. Van Der Meer, S. Fernandez, R. Kjelkenes, H. Ask, M. Rapp, M. Tschorn, S.J. Böttger, A. Marquand, G. Novarino, L. Marr, M. Slater, G.F. Viapiana, F.E. Orosa, J. Gallego, A. Pastor, A.J. Forstner, P. Hoffmann, M.M. Nöthen, I. Claus, A. Miller, C.M. Mathey, S. Heilmann-Heimbach, P. Sommer, M. Patraskaki, J. Wilbertz, K. Schmitt, V. Jirsa, S. Petkoski, S. Pitel, L. Otten, A.P. Athanasiadis, C. Pearmund, B. Spanlang, E. Alvarez, M. Sanchez, A. Giner, T. Jia, Y. Gong, Y. Xia, X. Chang, V. Calhoun, J. Liu, A. Schwalber, P. Thompson, N. Clinton, S. Desrivières, A.H. Young, B. Stahl, G. Ogoh, Nature Mental Health 2 (2024) 1124–1127."},"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","intvolume":"         2","status":"public","article_processing_charge":"No","OA_type":"closed access","issue":"10","day":"01","author":[{"first_name":"Frauke","full_name":"Nees, Frauke","last_name":"Nees"},{"first_name":"Paul","full_name":"Renner, Paul","last_name":"Renner"},{"first_name":"Nathalie E.","full_name":"Holz, Nathalie E.","last_name":"Holz"},{"first_name":"Elli","full_name":"Polemiti, Elli","last_name":"Polemiti"},{"first_name":"Sebastian","full_name":"Siehl, Sebastian","last_name":"Siehl"},{"full_name":"Hese, Sören","first_name":"Sören","last_name":"Hese"},{"first_name":"Kerstin","full_name":"Schepanski, Kerstin","last_name":"Schepanski"},{"last_name":"Schumann","full_name":"Schumann, Gunter","first_name":"Gunter"},{"last_name":"Walter","first_name":"Henrik","full_name":"Walter, Henrik"},{"last_name":"Heinz","first_name":"Andreas","full_name":"Heinz, Andreas"},{"last_name":"Ralser","full_name":"Ralser, Markus","first_name":"Markus"},{"last_name":"Twardziok","full_name":"Twardziok, Sven","first_name":"Sven"},{"last_name":"Vaidya","first_name":"Nilakshi","full_name":"Vaidya, Nilakshi"},{"last_name":"Bernas","first_name":"Antoine","full_name":"Bernas, Antoine"},{"last_name":"Serin","full_name":"Serin, Emin","first_name":"Emin"},{"full_name":"Jentsch, Marcel","first_name":"Marcel","last_name":"Jentsch"},{"full_name":"Hitchen, Esther","first_name":"Esther","last_name":"Hitchen"},{"full_name":"Kebir, Hedi","first_name":"Hedi","last_name":"Kebir"},{"last_name":"Lett","first_name":"Tristram A.","full_name":"Lett, Tristram A."},{"first_name":"Jean Charles","full_name":"Roy, Jean Charles","last_name":"Roy"},{"first_name":"Roland","full_name":"Eils, Roland","last_name":"Eils"},{"last_name":"Taron","first_name":"Ulrike Helene","full_name":"Taron, Ulrike Helene"},{"last_name":"Schütz","first_name":"Tatjana","full_name":"Schütz, Tatjana"},{"last_name":"Banks","full_name":"Banks, Jamie","first_name":"Jamie"},{"first_name":"Tobias","full_name":"Banaschewski, Tobias","last_name":"Banaschewski"},{"last_name":"Jansone","full_name":"Jansone, Karina","first_name":"Karina"},{"last_name":"Christmann","first_name":"Nina","full_name":"Christmann, Nina"},{"last_name":"Meyer-Lindenberg","full_name":"Meyer-Lindenberg, Andreas","first_name":"Andreas"},{"last_name":"Tost","first_name":"Heike","full_name":"Tost, Heike"},{"first_name":"Nathalie","full_name":"Holz, Nathalie","last_name":"Holz"},{"full_name":"Schwarz, Emanuel","first_name":"Emanuel","last_name":"Schwarz"},{"last_name":"Stringaris","first_name":"Argyris","full_name":"Stringaris, Argyris"},{"last_name":"Neidhart","full_name":"Neidhart, Maja","first_name":"Maja"},{"first_name":"Beke","full_name":"Seefried, Beke","last_name":"Seefried"},{"full_name":"Aden, Rieke","first_name":"Rieke","last_name":"Aden"},{"first_name":"Ole A.","full_name":"Andreassen, Ole A.","last_name":"Andreassen"},{"first_name":"Lars T.","full_name":"Westlye, Lars T.","last_name":"Westlye"},{"last_name":"Van Der Meer","full_name":"Van Der Meer, Dennis","first_name":"Dennis"},{"first_name":"Sara","full_name":"Fernandez, Sara","last_name":"Fernandez"},{"first_name":"Rikka","full_name":"Kjelkenes, Rikka","last_name":"Kjelkenes"},{"first_name":"Helga","full_name":"Ask, Helga","last_name":"Ask"},{"first_name":"Michael","full_name":"Rapp, Michael","last_name":"Rapp"},{"first_name":"Mira","full_name":"Tschorn, Mira","last_name":"Tschorn"},{"last_name":"Böttger","full_name":"Böttger, Sarah Jane","first_name":"Sarah Jane"},{"last_name":"Marquand","full_name":"Marquand, Andre","first_name":"Andre"},{"orcid":"0000-0002-7673-7178","id":"3E57A680-F248-11E8-B48F-1D18A9856A87","last_name":"Novarino","first_name":"Gaia","full_name":"Novarino, Gaia"},{"first_name":"Lena","full_name":"Marr, Lena","last_name":"Marr","id":"4406F586-F248-11E8-B48F-1D18A9856A87"},{"last_name":"Slater","full_name":"Slater, Mel","first_name":"Mel"},{"first_name":"Guillem Feixas","full_name":"Viapiana, Guillem Feixas","last_name":"Viapiana"},{"full_name":"Orosa, Francisco Eiroa","first_name":"Francisco Eiroa","last_name":"Orosa"},{"first_name":"Jaime","full_name":"Gallego, Jaime","last_name":"Gallego"},{"last_name":"Pastor","full_name":"Pastor, Alvaro","first_name":"Alvaro"},{"last_name":"Forstner","first_name":"Andreas J.","full_name":"Forstner, Andreas J."},{"first_name":"Per","full_name":"Hoffmann, Per","last_name":"Hoffmann"},{"last_name":"Nöthen","full_name":"Nöthen, Markus M.","first_name":"Markus M."},{"last_name":"Claus","full_name":"Claus, Isabelle","first_name":"Isabelle"},{"first_name":"Abigail","full_name":"Miller, Abigail","last_name":"Miller"},{"full_name":"Mathey, Carina M.","first_name":"Carina M.","last_name":"Mathey"},{"first_name":"Stefanie","full_name":"Heilmann-Heimbach, Stefanie","last_name":"Heilmann-Heimbach"},{"last_name":"Sommer","first_name":"Peter","full_name":"Sommer, Peter"},{"last_name":"Patraskaki","full_name":"Patraskaki, Myrto","first_name":"Myrto"},{"last_name":"Wilbertz","full_name":"Wilbertz, Johannes","first_name":"Johannes"},{"last_name":"Schmitt","full_name":"Schmitt, Karen","first_name":"Karen"},{"first_name":"Viktor","full_name":"Jirsa, Viktor","last_name":"Jirsa"},{"full_name":"Petkoski, Spase","first_name":"Spase","last_name":"Petkoski"},{"first_name":"Séverine","full_name":"Pitel, Séverine","last_name":"Pitel"},{"first_name":"Lisa","full_name":"Otten, Lisa","last_name":"Otten"},{"first_name":"Anastasios Polykarpos","full_name":"Athanasiadis, Anastasios Polykarpos","last_name":"Athanasiadis"},{"full_name":"Pearmund, Charlie","first_name":"Charlie","last_name":"Pearmund"},{"full_name":"Spanlang, Bernhard","first_name":"Bernhard","last_name":"Spanlang"},{"last_name":"Alvarez","first_name":"Elena","full_name":"Alvarez, Elena"},{"first_name":"Mavi","full_name":"Sanchez, Mavi","last_name":"Sanchez"},{"first_name":"Arantxa","full_name":"Giner, Arantxa","last_name":"Giner"},{"first_name":"Tianye","full_name":"Jia, Tianye","last_name":"Jia"},{"last_name":"Gong","full_name":"Gong, Yanting","first_name":"Yanting"},{"last_name":"Xia","full_name":"Xia, Yunman","first_name":"Yunman"},{"last_name":"Chang","full_name":"Chang, Xiao","first_name":"Xiao"},{"full_name":"Calhoun, Vince","first_name":"Vince","last_name":"Calhoun"},{"full_name":"Liu, Jingyu","first_name":"Jingyu","last_name":"Liu"},{"first_name":"Ameli","full_name":"Schwalber, Ameli","last_name":"Schwalber"},{"first_name":"Paul","full_name":"Thompson, Paul","last_name":"Thompson"},{"full_name":"Clinton, Nicholas","first_name":"Nicholas","last_name":"Clinton"},{"last_name":"Desrivières","full_name":"Desrivières, Sylvane","first_name":"Sylvane"},{"first_name":"Allan H.","full_name":"Young, Allan H.","last_name":"Young"},{"first_name":"Bernd","full_name":"Stahl, Bernd","last_name":"Stahl"},{"last_name":"Ogoh","full_name":"Ogoh, George","first_name":"George"}],"abstract":[{"lang":"eng","text":"This Comment explores new approaches to enrich large-scale population data, including incorporating macro-environmental and digital health measures."}],"title":"Large-scale population data enrichment in mental health research","page":"1124-1127","publication_status":"published","acknowledgement":"Funded by the European Union. Complementary funding was received by the UK Research and Innovation (UKRI) under the UK government’s Horizon Europe funding guarantee (10041392 and 10038599). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union, the European Health and Digital Executive Agency (HADEA) or UKRI. The European Union, HADEA and UKRI cannot be held responsible for them. This work received also support from Chinese Ministry for Science and Technology (MOST), the Horizon 2020-funded European Research Council Advanced Grant ‘STRATIFY’ (695313), the German Research Foundation (COPE; 675346; NE 1383/15-1 (CoviDrug)) and the National Natural Science Foundation of China grant 82150710554.","publication":"Nature Mental Health","publication_identifier":{"eissn":["2731-6076"]},"language":[{"iso":"eng"}],"article_type":"letter_note","year":"2024"},{"publication_status":"published","acknowledgement":"The authors thank Razvan Pascanu, Mahdi Nikdan and Soroush Tabesh for their valuable feedback, the IT department from Institute of Science and Technology Austria for the hardware support and Weights and Biases for the infrastructure to track all our experiments. Mher Safaryan has received funding from the European Union’s Horizon 2020 research and innovation program under the Marie Sklodowska-Curie grant agreement No 101034413.","abstract":[{"lang":"eng","text":"We propose a new variant of the Adam optimizer [Kingma and Ba, 2014] called\r\nMICROADAM that specifically minimizes memory overheads, while maintaining\r\ntheoretical convergence guarantees. We achieve this by compressing the gradient\r\ninformation before it is fed into the optimizer state, thereby reducing its memory\r\nfootprint significantly. We control the resulting compression error via a novel\r\ninstance of the classical error feedback mechanism from distributed optimization [Seide et al., 2014, Alistarh et al., 2018, Karimireddy et al., 2019] in which\r\nthe error correction information is itself compressed to allow for practical memory\r\ngains. We prove that the resulting approach maintains theoretical convergence\r\nguarantees competitive to those of AMSGrad, while providing good practical performance. Specifically, we show that MICROADAM can be implemented efficiently\r\non GPUs: on both million-scale (BERT) and billion-scale (LLaMA) models, MICROADAM provides practical convergence competitive to that of the uncompressed\r\nAdam baseline, with lower memory usage and similar running time. Our code is\r\navailable at https://github.com/IST-DASLab/MicroAdam."}],"title":"MICROADAM: Accurate adaptive optimization with low space overhead and provable convergence","author":[{"full_name":"Modoranu, Ionut-Vlad","first_name":"Ionut-Vlad","last_name":"Modoranu","id":"449f7a18-f128-11eb-9611-9b430c0c6333"},{"first_name":"Mher","full_name":"Safaryan, Mher","last_name":"Safaryan","id":"dd546b39-0804-11ed-9c55-ef075c39778d"},{"last_name":"Malinovsky","first_name":"Grigory","full_name":"Malinovsky, Grigory"},{"first_name":"Eldar","full_name":"Kurtic, Eldar","id":"47beb3a5-07b5-11eb-9b87-b108ec578218","last_name":"Kurtic"},{"id":"de632733-1457-11f0-ae22-b5914b8c1c41","last_name":"Robert","full_name":"Robert, Thomas","first_name":"Thomas"},{"last_name":"Richtárik","first_name":"Peter","full_name":"Richtárik, Peter"},{"first_name":"Dan-Adrian","full_name":"Alistarh, Dan-Adrian","id":"4A899BFC-F248-11E8-B48F-1D18A9856A87","last_name":"Alistarh","orcid":"0000-0003-3650-940X"}],"day":"20","OA_place":"repository","oa":1,"OA_type":"green","article_processing_charge":"No","year":"2024","external_id":{"arxiv":["2405.15593"]},"language":[{"iso":"eng"}],"publication_identifier":{"issn":["1049-5258"]},"acknowledged_ssus":[{"_id":"CampIT"}],"publication":"38th Conference on Neural Information Processing Systems","main_file_link":[{"url":"https://doi.org/10.48550/arXiv.2405.15593","open_access":"1"}],"_id":"19510","oa_version":"Preprint","month":"12","date_created":"2025-04-06T22:01:32Z","volume":37,"quality_controlled":"1","date_updated":"2025-05-14T11:32:52Z","scopus_import":"1","project":[{"call_identifier":"H2020","grant_number":"101034413","_id":"fc2ed2f7-9c52-11eb-aca3-c01059dda49c","name":"IST-BRIDGE: International postdoctoral program"}],"related_material":{"link":[{"relation":"software","url":"https://github.com/IST-DASLab/MicroAdam"}]},"type":"conference","corr_author":"1","status":"public","ec_funded":1,"arxiv":1,"alternative_title":["Advances in Neural Information Processing Systems"],"date_published":"2024-12-20T00:00:00Z","citation":{"ista":"Modoranu I-V, Safaryan M, Malinovsky G, Kurtic E, Robert T, Richtárik P, Alistarh D-A. 2024. MICROADAM: Accurate adaptive optimization with low space overhead and provable convergence. 38th Conference on Neural Information Processing Systems. , Advances in Neural Information Processing Systems, vol. 37.","mla":"Modoranu, Ionut-Vlad, et al. “MICROADAM: Accurate Adaptive Optimization with Low Space Overhead and Provable Convergence.” <i>38th Conference on Neural Information Processing Systems</i>, vol. 37, Neural Information Processing Systems Foundation, 2024.","chicago":"Modoranu, Ionut-Vlad, Mher Safaryan, Grigory Malinovsky, Eldar Kurtic, Thomas Robert, Peter Richtárik, and Dan-Adrian Alistarh. “MICROADAM: Accurate Adaptive Optimization with Low Space Overhead and Provable Convergence.” In <i>38th Conference on Neural Information Processing Systems</i>, Vol. 37. Neural Information Processing Systems Foundation, 2024.","short":"I.-V. Modoranu, M. Safaryan, G. Malinovsky, E. Kurtic, T. Robert, P. Richtárik, D.-A. Alistarh, in:, 38th Conference on Neural Information Processing Systems, Neural Information Processing Systems Foundation, 2024.","ama":"Modoranu I-V, Safaryan M, Malinovsky G, et al. MICROADAM: Accurate adaptive optimization with low space overhead and provable convergence. In: <i>38th Conference on Neural Information Processing Systems</i>. Vol 37. Neural Information Processing Systems Foundation; 2024.","apa":"Modoranu, I.-V., Safaryan, M., Malinovsky, G., Kurtic, E., Robert, T., Richtárik, P., &#38; Alistarh, D.-A. (2024). MICROADAM: Accurate adaptive optimization with low space overhead and provable convergence. In <i>38th Conference on Neural Information Processing Systems</i> (Vol. 37). Neural Information Processing Systems Foundation.","ieee":"I.-V. Modoranu <i>et al.</i>, “MICROADAM: Accurate adaptive optimization with low space overhead and provable convergence,” in <i>38th Conference on Neural Information Processing Systems</i>, 2024, vol. 37."},"intvolume":"        37","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","publisher":"Neural Information Processing Systems Foundation","department":[{"_id":"DaAl"}]}]