@misc{19884,
  abstract     = {This is Marlin, a Mixed Auto-Regressive Linear kernel (and the name of one of the planet's fastest fish), an extremely optimized FP16xINT4 matmul kernel aimed at LLM inference that can deliver close to ideal (4x) speedups up to batchsizes of 16-32 tokens (in contrast to the 1-2 tokens of prior work with comparable speedup).

Additionally, it includes Sparse-Marlin, an extension of the MARLIN kernels adding support to 2:4 weight sparsity, achieving 5.3x speedups on NVIDIA GPUs (Ampere/Ada).},
  author       = {Frantar, Elias and Castro, Roberto and Chen, Jiale and Hoefler, Torsten and Alistarh, Dan-Adrian},
  publisher    = {Zenodo},
  title        = {{MARLIN: Mixed-precision auto-regressive parallel inference on Large Language Models}},
  doi          = {10.5281/ZENODO.14213091},
  year         = {2024},
}

@article{20039,
  abstract     = {This Comment presents a high-level protocol for data harmonization within large cohorts, in which it postulates four main steps including (1) expert review, (2) pre-statistical harmonization, (3) statistical harmonization, and (4) validation.},
  author       = {Neidhart, Maja and Kjelkenes, Rikka and Jansone, Karina and Rehák Bučková, Barbora and Holz, Nathalie and Nees, Frauke and Walter, Henrik and Schumann, Gunter and Rapp, Michael A. and Banaschewski, Tobias and Schwarz, Emanuel and Marquand, Andre and Ogoh, George and Stahl, Bernd and Young, Allan H. and Desrivières, Sylvane and Clinton, Nicholas and Thompson, Paul and Schwalber, Ameli and Liu, Jingyu and Calhoun, Vince and Chang, Xiao and Xia, Yunman and Gong, Yanting and Jia, Tianye and Renner, Paul and Hese, Sören and Giner, Arantxa and Sanchez, Mavi and Alvarez, Elena and Spanlang, Bernhard and Pearmund, Charlie and Athanasiadis, Anastasios Polykarpos and Otten, Lisa and Pitel, Séverine and Petkoski, Spase and Jirsa, Viktor and Schmitt, Karen and Wilbertz, Johannes and Patraskaki, Myrto and Sommer, Peter and Heilmann-Heimbach, Stefanie and Mathey, Carina M. and Miller, Abigail and Claus, Isabelle and Nöthen, Markus M. and Hoffmann, Per and Forstner, Andreas J. and Pastor, Alvaro and Gallego, Jaime and Orosa, Francisco Eiroa and Viapiana, Guillem Feixas and Slater, Mel and Marr, Lena and Novarino, Gaia and Böttger, Sarah Jane and Tschorn, Mira and Rapp, Michael and Ask, Helga and Fernandez, Sara and Van Der Meer, Dennis and Westlye, Lars T. and Andreassen, Ole A. and Aden, Rieke and Seefried, Beke and Siehl, Sebastian and Nees, Frauke and Stringaris, Argyris and Tost, Heike and Meyer-Lindenberg, Andreas and Christmann, Nina and Banks, Jamie and Schepanski, Kerstin and Schütz, Tatjana and Taron, Ulrike Helene and Eils, Roland and Roy, Jean Charles and Lett, Tristram A. and Kebir, Hedi and Polemiti, Elli and Hitchen, Esther and Jentsch, Marcel and Serin, Emin and Bernas, Antoine and Vaidya, Nilakshi and Twardziok, Sven and Ralser, Markus and Heinz, Andreas},
  issn         = {2731-6076},
  journal      = {Nature Mental Health},
  number       = {10},
  pages        = {1134--1137},
  publisher    = {Springer Nature},
  title        = {{A protocol for data harmonization in large cohorts}},
  doi          = {10.1038/s44220-024-00315-0},
  volume       = {2},
  year         = {2024},
}

@misc{20057,
  abstract     = {In article number 2305128, Qing Sun, Shang Wang, Yanhong Tian, Andreu Cabot, and co-workers report an investigation of the energy-storage mechanism of a layered Bi2Te3-based cathode for aqueous zinc-ion batteries (ZIBs). They demonstrate that the zinc ion is not inserted into the cathode as previously assumed; in contrast, proton charge-storage dominates the process. They also demonstrate the great application prospects of aqueous ZIBs in flexible electronics via jet printing technology.},
  author       = {Zeng, Guifang and Sun, Qing and Horta, Sharona and Wang, Shang and Lu, Xuan and Zhang, Chao Yue and Li, Jing and Li, Junshan and Ci, Lijie and Tian, Yanhong and Ibáñez, Maria and Cabot, Andreu},
  booktitle    = {Advanced Materials},
  issn         = {1521-4095},
  number       = {1},
  publisher    = {Wiley},
  title        = {{A layered Bi2Te3@PPy cathode for aqueous Zinc‐Ion batteries: Mechanism and application in printed flexible batteries}},
  doi          = {10.1002/adma.202470004},
  volume       = {36},
  year         = {2024},
}

@misc{20121,
  abstract     = {PyDaddy is an open source package which is a key contribution of the manuscript Nabeel et al, arXiv:2205.02645. The basic scientific premise for this package is to discover the nature of stochasticity in ecological time series datasets. It is well known that the stochasticity can affect the dynamics of ecological systems in counter-intuitive ways. Without understanding the equations (typically, in the form of stochastic differential equations or SDEs, in short) that govern the dynamics of populations or ecosystems, it's challenging to determine the impact of randomness on real datasets. In this manuscript and accompanying package, we introduce a methodology for discovering equations (SDEs) that transforms time series data of state variables into stochastic differential equations. This approach merges traditional stochastic calculus with modern equation-discovery techniques. We showcase the generality of our method through various applications and discuss its limitations and potential pitfalls, offering diagnostic measures to address these challenges.},
  author       = {Nabeel, Arshed and Karichannavar, Ashwin and Palathingal, Shuaib and Jhawar, Jitesh and Brückner, David and Danny Raj, Masila and Guttal, Vishwesha},
  publisher    = {Zenodo},
  title        = {{PyDaddy: A Python Package for Discovering SDEs from Time Series Data}},
  doi          = {10.5281/ZENODO.7137151},
  year         = {2024},
}

@article{20156,
  abstract     = {Integrative analyses that incorporate different levels of ‘-omics’ data represent a powerful tool for deciphering the biological mechanisms that underlie environmental influences on mental health and disease. This Comment highlights various aspects of such multi-omics approaches, using the example of the EU-funded environMENTAL project.},
  author       = {Desrivières, Sylvane and Miller, Abigail and Mathey, Carina M. and Yu, Xinyang and Chen, Di and Agunbiade, Kofoworola and Heilmann-Heimbach, Stefanie and Forstner, Andreas J. and Schumann, Gunter and Hoffmann, Per and Nöthen, Markus M. and Ogoh, George and Stahl, Bernd and Young, Allan H. and Clinton, Nicholas and Thompson, Paul and Schwalber, Ameli and Liu, Jingyu and Calhoun, Vince and Chang, Xiao and Xia, Yunman and Gong, Yanting and Jia, Tianye and Renner, Paul and Hese, Sören and Giner, Arantxa and Sanchez, Mavi and Alvarez, Elena and Spanlang, Bernhard and Pearmund, Charlie and Athanasiadis, Anastasios Polykarpos and Otten, Lisa and Pitel, Séverine and Petkoski, Spase and Jirsa, Viktor and Schmitt, Karen and Wilbertz, Johannes and Patraskaki, Myrto and Sommer, Peter and Claus, Isabelle and Pastor, Alvaro and Gallego, Jaime and Orosa, Francisco Eiroa and Viapiana, Guillem Feixas and Slater, Mel and Marr, Lena and Novarino, Gaia and Marquand, Andre and Böttger, Sarah Jane and Tschorn, Mira and Rapp, Michael and Ask, Helga and Kjelkenes, Rikka and Fernandez, Sara and Van Der Meer, Dennis and Westlye, Lars T. and Andreassen, Ole A. and Aden, Rieke and Seefried, Beke and Siehl, Sebastian and Nees, Frauke and Neidhart, Maja and Stringaris, Argyris and Schwarz, Emanuel and Holz, Nathalie and Tost, Heike and Meyer-Lindenberg, Andreas and Christmann, Nina and Jansone, Karina and Banaschewski, Tobias and Banks, Jamie and Schepanski, Kerstin and Schütz, Tatjana and Taron, Ulrike Helene and Eils, Roland and Roy, Jean Charles and Lett, Tristram A. and Kebir, Hedi and Polemiti, Elli and Hitchen, Esther and Jentsch, Marcel and Serin, Emin and Bernas, Antoine and Vaidya, Nilakshi and Twardziok, Sven and Ralser, Markus and Heinz, Andreas and Walter, Henrik},
  issn         = {2731-6076},
  journal      = {Nature Mental Health},
  number       = {10},
  pages        = {1131--1133},
  publisher    = {Springer Nature},
  title        = {{Multi-omics analyses of the environMENTAL project provide insights into mental health and disease}},
  doi          = {10.1038/s44220-024-00317-y},
  volume       = {2},
  year         = {2024},
}

@article{20157,
  abstract     = {The focus of much of contemporary research ethics is on compliance with established protocols. However, large data-driven neuroscience research raises new ethical concerns that have no agreed-upon solution. Here we reflect on these challenges and propose better integration of public and patient involvement in this evolving landscape.},
  author       = {Stahl, Bernd and Ogoh, George and Schumann, Gunter and Walter, Henrik and Stahl, Bernd and Young, Allan H. and Desrivières, Sylvane and Clinton, Nicholas and Thompson, Paul and Schwalber, Ameli and Liu, Jingyu and Calhoun, Vince and Chang, Xiao and Xia, Yunman and Gong, Yanting and Jia, Tianye and Renner, Paul and Hese, Sören and Giner, Arantxa and Sanchez, Mavi and Alvarez, Elena and Spanlang, Bernhard and Pearmund, Charlie and Athanasiadis, Anastasios Polykarpos and Otten, Lisa and Pitel, Séverine and Petkoski, Spase and Jirsa, Viktor and Schmitt, Karen and Wilbertz, Johannes and Patraskaki, Myrto and Sommer, Peter and Heilmann-Heimbach, Stefanie and Mathey, Carina M. and Miller, Abigail and Claus, Isabelle and Nöthen, Markus M. and Hoffmann, Per and Forstner, Andreas J. and Pastor, Alvaro and Gallego, Jaime and Orosa, Francisco Eiroa and Viapiana, Guillem Feixas and Slater, Mel and Marr, Lena and Novarino, Gaia and Marquand, Andre and Böttger, Sarah Jane and Tschorn, Mira and Rapp, Michael and Ask, Helga and Kjelkenes, Rikka and Fernandez, Sara and Van Der Meer, Dennis and Westlye, Lars T. and Andreassen, Ole A. and Aden, Rieke and Seefried, Beke and Siehl, Sebastian and Nees, Frauke and Neidhart, Maja and Stringaris, Argyris and Schwarz, Emanuel and Holz, Nathalie and Tost, Heike and Meyer-Lindenberg, Andreas and Christmann, Nina and Jansone, Karina and Banaschewski, Tobias and Banks, Jamie and Schepanski, Kerstin and Schütz, Tatjana and Taron, Ulrike Helene and Eils, Roland and Roy, Jean Charles and Lett, Tristram A. and Kebir, Hedi and Polemiti, Elli and Hitchen, Esther and Jentsch, Marcel and Serin, Emin and Bernas, Antoine and Vaidya, Nilakshi and Twardziok, Sven and Ralser, Markus and Heinz, Andreas and Walter, Henrik},
  issn         = {2731-6076},
  journal      = {Nature Mental Health},
  number       = {10},
  publisher    = {Springer Nature},
  title        = {{Rethinking ethics in interdisciplinary and big data-driven neuroscience projects}},
  doi          = {10.1038/s44220-024-00320-3},
  volume       = {2},
  year         = {2024},
}

@unpublished{20701,
  abstract     = {A Proof of Exponentiation (PoE) allows a prover to efficiently convince a verifier that 𝑦 = 𝑥
𝑒
in some group of unknown order. PoEs
are the basis for practical constructions of Verifiable Delay Functions (VDFs), which, in turn, are important for various higher-level
protocols in distributed computing. In applications such as distributed consensus, many PoEs are generated regularly, motivating
protocols for secure aggregation of batches of statements into a
few statements to improve the efficiency for both parties. Rotem
(TCC 2021) recently presented two such generic batch PoEs.
In this work, we introduce two batch PoEs that outperform
both proposals of Rotem and we evaluate their practicality. First,
we show that the two batch PoEs of Rotem can be combined to
improve the overall efficiency by at least a factor of two. Second, we
revisit the work of Bellare, Garay, and Rabin (EUROCRYPT 1998)
on batch verification of digital signatures and show that, under the
low order assumption, their bucket test can be securely adapted to
the setting of groups of unknown order. The resulting batch PoE
quickly outperforms the state of the art in the expected number of
group multiplications with the growing number of instances, and it
decreases the cost of batching by an order of magnitude already for
hundreds of thousands of instances. Importantly, it is the first batch
PoE that significantly decreases both the proof size and complexity
of verification. Our experimental evaluations show that even a nonoptimized implementation achieves such improvements, which
would match the demands of real-life systems requiring large-scale
PoE processing.
Finally, even though our proof techniques are conceptually similar to Rotem, we give an improved analysis of the application of the
low order assumption towards secure batching of PoE instances,
resulting in a tight reduction, which is important when setting the
security parameter in practice.},
  author       = {Hoffmann, Charlotte and Hubáček, Pavel and Ivanova, Svetlana},
  booktitle    = {Cryptology ePrint Archive},
  publisher    = {International Association for Cryptologic Research },
  title        = {{Practical batch proofs of exponentiation}},
  year         = {2024},
}

@inproceedings{18755,
  abstract     = {A universalthresholdizer (UT), constructed from a threshold fully homomorphic encryption by Boneh et. al , Crypto 2018, is a general framework for universally thresholdizing many cryptographic schemes. However, their framework is insufficient to construct strongly secure threshold schemes, such as threshold signatures and threshold public-key encryption, etc.

In this paper, we strengthen the security definition for a universal thresholdizer and propose a scheme which satisfies our stronger security notion. Our UT scheme is an improvement of Boneh et. al ’s construction at the level of threshold fully homomorphic encryption using a key homomorphic pseudorandom function. We apply our strongly secure UT scheme to construct strongly secure threshold signatures and threshold public-key encryption.},
  author       = {Ebrahimi, Ehsan and Yadav, Anshu},
  booktitle    = {30th International Conference on the Theory and Application of Cryptology and Information Security},
  isbn         = {9789819608904},
  issn         = {1611-3349},
  location     = {Kolkata, India},
  pages        = {207--239},
  publisher    = {Springer Nature},
  title        = {{Strongly secure universal thresholdizer}},
  doi          = {10.1007/978-981-96-0891-1_7},
  volume       = {15486},
  year         = {2024},
}

@inproceedings{18756,
  abstract     = {The evasive LWE assumption, proposed by Wee [Eurocrypt’22 Wee] for constructing a lattice-based optimal broadcast encryption, has shown to be a powerful assumption, adopted by subsequent works to construct advanced primitives ranging from ABE variants to obfuscation for null circuits. However, a closer look reveals significant differences among the precise assumption statements involved in different works, leading to the fundamental question of how these assumptions compare to each other. In this work, we initiate a more systematic study on evasive LWE assumptions:
(i) Based on the standard LWE assumption, we construct simple counterexamples against three private-coin evasive LWE variants, used in [Crypto’22 Tsabary, Asiacrypt’22 VWW, Crypto’23 ARYY] respectively, showing that these assumptions are unlikely to hold.

(ii) Based on existing evasive LWE variants and our counterexamples, we propose and define three classes of plausible evasive LWE assumptions, suitably capturing all existing variants for which we are not aware of non-obfuscation-based counterexamples.

(iii) We show that under our assumption formulations, the security proofs of [Asiacrypt’22 VWW] and [Crypto’23 ARYY] can be recovered, and we reason why the security proof of [Crypto’22 Tsabary] is also plausibly repairable using an appropriate evasive LWE assumption.},
  author       = {Brzuska, Chris and Ünal, Akin and Woo, Ivy K.Y.},
  booktitle    = {30th International Conference on the Theory and Application of Cryptology and Information Security},
  isbn         = {9789819608935},
  issn         = {1611-3349},
  location     = {Kolkata, India},
  pages        = {418--449},
  publisher    = {Springer Nature},
  title        = {{Evasive LWE assumptions: Definitions, classes, and counterexamples}},
  doi          = {10.1007/978-981-96-0894-2_14},
  volume       = {15487},
  year         = {2024},
}

@article{18757,
  abstract     = {Segmentation is a critical data processing step in many applications of cryo-electron tomography. Downstream analyses, such as subtomogram averaging, are often based on segmentation results, and are thus critically dependent on the availability of open-source software for accurate as well as high-throughput tomogram segmentation. There is a need for more user-friendly, flexible, and comprehensive segmentation software that offers an insightful overview of all steps involved in preparing automated segmentations. Here, we present Ais: a dedicated tomogram segmentation package that is geared towards both high performance and accessibility, available on GitHub. In this report, we demonstrate two common processing steps that can be greatly accelerated with Ais: particle picking for subtomogram averaging, and generating many-feature segmentations of cellular architecture based on in situ tomography data. Featuring comprehensive annotation, segmentation, and rendering functionality, as well as an open repository for trained models at aiscryoet.org, we hope that Ais will help accelerate research and dissemination of data involving cryoET.},
  author       = {Last, Mart G.F. and Abendstein, Leoni and Voortman, Lenard M. and Sharp, Thomas H.},
  issn         = {2050-084X},
  journal      = {eLife},
  publisher    = {eLife Sciences Publications},
  title        = {{Streamlining segmentation of cryo-electron tomography datasets with Ais}},
  doi          = {10.7554/eLife.98552},
  volume       = {13},
  year         = {2024},
}

@inproceedings{18758,
  abstract     = {MaxCut is a classical NP-complete problem and a crucial building block in many combinatorial algorithms. The famous Edwards-Erdős bound states that any connected graph on n vertices with m edges contains a cut of size at least m/2+(n-1)/4. Crowston, Jones and Mnich [Algorithmica, 2015] showed that the MaxCut problem on simple connected graphs admits an FPT algorithm, where the parameter k is the difference between the desired cut size c and the lower bound given by the Edwards-Erdős bound. This was later improved by Etscheid and Mnich [Algorithmica, 2017] to run in parameterized linear time, i.e., f(k)⋅ O(m). We improve upon this result in two ways: Firstly, we extend the algorithm to work also for multigraphs (alternatively, graphs with positive integer weights). Secondly, we change the parameter; instead of the difference to the Edwards-Erdős bound, we use the difference to the Poljak-Turzík bound. The Poljak-Turzík bound states that any weighted graph G has a cut of size at least (w(G))/2+(w_MSF(G))/4, where w(G) denotes the total weight of G, and w_MSF(G) denotes the weight of its minimum spanning forest. In connected simple graphs the two bounds are equivalent, but for multigraphs the Poljak-Turzík bound can be larger and thus yield a smaller parameter k. Our algorithm also runs in parameterized linear time, i.e., f(k)⋅ O(m+n).},
  author       = {Lill, Jonas and Petrova, Kalina H and Weber, Simon},
  booktitle    = {19th International Symposium on Parameterized and Exact Computation},
  isbn         = {9783959773539},
  issn         = {1868-8969},
  location     = {Egham, United Kingdom},
  publisher    = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik},
  title        = {{Linear-time MaxCut in multigraphs parameterized above the Poljak-Turzík bound}},
  doi          = {10.4230/LIPIcs.IPEC.2024.2},
  volume       = {321},
  year         = {2024},
}

@article{18760,
  abstract     = {With the remarkable sensitivity and resolution of JWST in the infrared, measuring rest-optical kinematics of galaxies at z > 5 has become possible for the first time. This study pilots a new method for measuring galaxy dynamics for highly multiplexed, unbiased samples by combining FRESCO NIRCam grism spectroscopy and JADES medium-band imaging. Here we present one of the first JWST kinematic measurements for a galaxy at z > 5. We find a significant velocity gradient, which, if interpreted as rotation, yields Vrot = 305 ± 70 km s−1, and we hence refer to this galaxy as Twister-z5. With a rest-frame optical effective radius of re = 2.25 kpc, the high rotation velocity in this galaxy is not due to a compact size, as may be expected in the early Universe, but rather to a high total mass, (math formula). This is a factor of roughly 10× higher than the stellar mass within re. We also observe that the radial Hα equivalent width profile and the specific star formation rate map from resolved stellar population modeling are centrally depressed by a factor of ∼1.5 from the center to re. Combined with the morphology of the line-emitting gas in comparison to the continuum, this centrally suppressed star formation is consistent with a star-forming disk surrounding a bulge growing inside out. While large, rapidly rotating disks are common to z ∼ 2, the existence of one after only 1 Gyr of cosmic time, shown for the first time in ionized gas, adds to the growing evidence that some galaxies matured earlier than expected in the history of the Universe.},
  author       = {Nelson, Erica and Brammer, Gabriel and Giménez-Arteaga, Clara and Oesch, Pascal A. and Naidu, Rohan P. and Übler, Hannah and Matharu, Jasleen and Shapley, Alice E. and Whitaker, Katherine E. and Wisnioski, Emily and Förster Schreiber, Natascha M. and Smit, Renske and Van Dokkum, Pieter and Chisholm, John and Endsley, Ryan and Hartley, Abigail I. and Gibson, Justus and Giovinazzo, Emma and Illingworth, Garth and Labbe, Ivo and Maseda, Michael V. and Matthee, Jorryt J and Covelo Paz, Alba and Price, Sedona H. and Reddy, Naveen A. and Shivaei, Irene and Weibel, Andrea and Wuyts, Stijn and Xiao, Mengyuan and Alberts, Stacey and Baker, William M. and Bunker, Andrew J. and Cameron, Alex J. and Charlot, Stephane and Eisenstein, Daniel J. and De Graaff, Anna and Ji, Zhiyuan and Johnson, Benjamin D. and Jones, Gareth C. and Maiolino, Roberto and Robertson, Brant and Sandles, Lester and Suess, Katherine A. and Tacchella, Sandro and Williams, Christina C. and Witstok, Joris},
  issn         = {2041-8213},
  journal      = {Astrophysical Journal Letters},
  number       = {2},
  publisher    = {IOP Publishing},
  title        = {{Ionized gas kinematics with FRESCO: An extended, massive, rapidly rotating galaxy at z = 5.4}},
  doi          = {10.3847/2041-8213/ad7b17},
  volume       = {976},
  year         = {2024},
}

@article{18761,
  abstract     = {Termites, together with cockroaches, belong to the Blattodea. They possess an XX/XY sex determination system which has evolved from an XX/X0 system present in other Blattodean species, such as cockroaches and wood roaches. Little is currently known about the sex chromosomes of termites, their gene content, or their evolution. We here investigate the X chromosome of multiple termite species and compare them with the X chromosome of cockroaches using genomic and transcriptomic data. We find that the X chromosome of the termite Macrotermes natalensis is large and differentiated showing hall marks of sex chromosome evolution such as dosage compensation, while this does not seem to be the case in the other two termite species investigated here where sex chromosomes may be evolutionary younger. Furthermore, the X chromosome in M. natalensis is different from the X chromosome found in the cockroach Blattella germanica indicating that sex chromosome turn-over events may have happened during termite evolution.},
  author       = {Fraser, Roxanne and Moraa, Ruth and Djolai, Annika and Meisenheimer, Nils and Laube, Sophie and Vicoso, Beatriz and Huylmans, Ann K},
  issn         = {1759-6653},
  journal      = {Genome Biology and Evolution},
  number       = {12},
  publisher    = {Oxford University Press},
  title        = {{Evidence for a novel X chromosome in termites}},
  doi          = {10.1093/gbe/evae265},
  volume       = {16},
  year         = {2024},
}

@article{18762,
  abstract     = {Consider the random variable $\mathrm{Tr}( f_1(W)A_1\dots f_k(W)A_k)$ where $W$ is an $N\times N$ Hermitian Wigner matrix, $k\in\mathbb{N}$, and choose (possibly $N$-dependent) regular functions $f_1,\dots, f_k$ as well as bounded deterministic matrices $A_1,\dots,A_k$. We give a functional central limit theorem showing that the fluctuations around the expectation are Gaussian. Moreover, we determine the limiting covariance structure and give explicit error bounds in terms of the scaling of $f_1,\dots,f_k$ and the number of traceless matrices among $A_1,\dots,A_k$, thus extending the results of [Cipolloni, Erdős, Schröder 2023] to products of arbitrary length $k\geq2$. As an application, we consider the fluctuation of $\mathrm{Tr}(\mathrm{e}^{\mathrm{i} tW}A_1\mathrm{e}^{-\mathrm{i} tW}A_2)$ around its thermal value $\mathrm{Tr}(A_1)\mathrm{Tr}(A_2)$ when $t$ is large and give an explicit formula for the variance.},
  author       = {Reker, Jana},
  issn         = {1083-6489},
  journal      = {Electronic Journal of Probability},
  publisher    = {Institute of Mathematical Statistics},
  title        = {{Multi-point functional central limit theorem for Wigner matrices}},
  doi          = {10.1214/24-EJP1247},
  volume       = {29},
  year         = {2024},
}

@phdthesis{18766,
  abstract     = {Poxviruses are large pleomorphic double-stranded DNA viruses that include well known members such as variola virus, the causative agent of smallpox, Mpox virus, as well as Vaccinia virus (VACV), which serves as a vaccination strain for formerly mentioned viruses. VACV is a valuable model for studying large pleomorphic DNA viruses in general and poxviruses specifically, as many features, such as core morphology and structural proteins, are well conserved within this family. Despite decades of research, our understanding of the structural components and proteins that comprise the poxvirus core in mature virions remains limited. Although major core proteins were identified via indirect experimental evidence, the core's complexity, with its large size, structure and number of involved proteins, has hindered efforts to achieve high-resolution insights and to define the roles of the individual proteins. The specific protein composition of the core's individual layers, including the palisade layer and the inner core wall, has remained unclear. In this study, we have merged multiple approaches, including single particle cryo electron microscopy of purified virus cores, cryo-electron tomography and subtomogram averaging of mature virions and molecular modeling to elucidate the structural determinants of the VACV core. Due to the lack of experimentally derived structures, either in situ or reconstituted in vitro, we used Alphafold to predict models of the putative major core protein candidates, A10, 23k, A3, A4, and L4. Our results show that the VACV core is composed of several layers with varying local symmetries, forming more intricate interactions than observed previously. This allowed us to identify several molecular building blocks forming the viral core lattice. In particular, we identified trimers of protein A10 as a major core structure that forms the palisade layer of the viral core. Additionally, we revealed that six petals of a flower shaped core pore within the core wall are composed of A10 trimers. Furthermore, we obtained a cryo-EM density for the inner core wall that could potentially accommodate an A3 dimer. Integrating descriptions of protein interactions from previous studies enabled us to provide a detailed structural model of the poxvirus core wall, and our findings indicate that the interactions within A10 trimers are likely consistent across orthopox- and parapoxviruses. This combined application of cryo-SPA and cryo-ET can help overcome obstacles in studying complex virus structures in the future, including their key assembly proteins, interactions, and the formation into a core lattice. Our work provides important fundamental new insights into poxvirus core architecture, also considering the recent re-emergence of poxviruses.},
  author       = {Datler, Julia},
  isbn         = {978-3-99078-049-7},
  issn         = {2663-337X},
  keywords     = {cryo-EM, cryo-ET, cryo-SPA, Structural Virology, Poxvirus, Vaccinia Virus, Structural Biology},
  pages        = {106},
  publisher    = {Institute of Science and Technology Austria},
  title        = {{Elucidating the structural determinants of the poxvirus core using multi-modal cryo-EM}},
  doi          = {10.15479/at:ista:18766},
  year         = {2024},
}

@article{18779,
  abstract     = {Unsupervised segmentation in biological and non-biological images is only partially resolved. Segmentation either requires arbitrary thresholds or large teaching datasets. Here, we propose a spatial autocorrelation method based on Local Moran’s <jats:italic>I</jats:italic> coefficient to differentiate signal, background, and noise in any type of image. The method, originally described for geoinformatics, does not require a predefined intensity threshold or teaching algorithm for image segmentation and allows quantitative comparison of samples obtained in different conditions. It utilizes relative intensity as well as spatial information of neighboring elements to select spatially contiguous groups of pixels. We demonstrate that Moran’s method outperforms threshold-based method in both artificially generated as well as in natural images especially when background noise is substantial. This superior performance can be attributed to the exclusion of false positive pixels resulting from isolated, high intensity pixels in high noise conditions. To test the method’s power in real situation, we used high power confocal images of the somatosensory thalamus immunostained for Kv4.2 and Kv4.3 (A-type) voltage-gated potassium channels in mice. Moran’s method identified high-intensity Kv4.2 and Kv4.3 ion channel clusters in the thalamic neuropil. Spatial distribution of these clusters displayed strong correlation with large sensory axon terminals of subcortical origin. The unique association of the special presynaptic terminals and a postsynaptic voltage-gated ion channel cluster was confirmed with electron microscopy. These data demonstrate that Moran’s method is a rapid, simple image segmentation method optimal for variable and high noise conditions.},
  author       = {Dávid, Csaba and Giber, Kristóf and Szigeti, Margit Katalin and Köllő, Mihály and Nusser, Zoltan and Acsady, Laszlo},
  issn         = {2050-084X},
  journal      = {eLife},
  publisher    = {eLife Sciences Publications},
  title        = {{A novel image segmentation method based on spatial autocorrelation identifies A-type potassium channel clusters in the thalamus}},
  doi          = {10.7554/elife.89361},
  volume       = {12},
  year         = {2024},
}

@inproceedings{18847,
  abstract     = {Machine Learning and AI have the potential to transform data-driven
scientific discovery, enabling accurate predictions for several scientific
phenomena. As many scientific questions are inherently causal, this paper looks
at the causal inference task of treatment effect estimation, where the outcome
of interest is recorded in high-dimensional observations in a Randomized
Controlled Trial (RCT). Despite being the simplest possible causal setting and
a perfect fit for deep learning, we theoretically find that many common choices
in the literature may lead to biased estimates. To test the practical impact of
these considerations, we recorded ISTAnt, the first real-world benchmark for
causal inference downstream tasks on high-dimensional observations as an RCT
studying how garden ants (Lasius neglectus) respond to microparticles applied
onto their colony members by hygienic grooming. Comparing 6 480 models
fine-tuned from state-of-the-art visual backbones, we find that the sampling
and modeling choices significantly affect the accuracy of the causal estimate,
and that classification accuracy is not a proxy thereof. We further validated
the analysis, repeating it on a synthetically generated visual data set
controlling the causal model. Our results suggest that future benchmarks should
carefully consider real downstream scientific questions, especially causal
ones. Further, we highlight guidelines for representation learning methods to
help answer causal questions in the sciences.},
  author       = {Cadei, Riccardo and Lindorfer, Lukas and Cremer, Sylvia and Schmid, Cordelia and Locatello, Francesco},
  booktitle    = {ICML 2024 Workshop AI4Science},
  publisher    = {Curran Associates},
  title        = {{Smoke and mirrors in causal downstream tasks}},
  volume       = {38},
  year         = {2024},
}

@article{18856,
  abstract     = {This research is aimed to solve the tweet/user geolocation prediction task and provide a flexible methodology for the geo-tagging of textual big data. The suggested approach implements neural networks for natural language processing (NLP) to estimate the location as coordinate pairs (longitude, latitude) and two-dimensional Gaussian Mixture Models (GMMs). The scope of proposed models has been finetuned on a Twitter dataset using pretrained Bidirectional Encoder Representations from Transformers (BERT) as base models. Performance metrics show a median error of fewer than 30 km on a worldwide-level, and fewer than 15 km on the US-level datasets for the models trained and evaluated on text features of tweets' content and metadata context. Our source code and data are available at https://github.com/K4TEL/geo-twitter.git.},
  author       = {Lutsai, Kateryna and Lampert, Christoph},
  issn         = {1948-660X},
  journal      = {Journal of Spatial Information Science},
  number       = {29},
  pages        = {69--99},
  publisher    = {University of Maine},
  title        = {{Predicting the geolocation of tweets using transformer models on customized data}},
  doi          = {10.5311/JOSIS.2024.29.295},
  year         = {2024},
}

@unpublished{18874,
  abstract     = {Despite extensive research since the community learned about adversarial
examples 10 years ago, we still do not know how to train high-accuracy
classifiers that are guaranteed to be robust to small perturbations of their
inputs. Previous works often argued that this might be because no classifier
exists that is robust and accurate at the same time. However, in computer
vision this assumption does not match reality where humans are usually accurate
and robust on most tasks of interest. We offer an alternative explanation and
show that in certain settings robust generalization is only possible with
unrealistically large amounts of data. More precisely we find a setting where a
robust classifier exists, it is easy to learn an accurate classifier, yet it
requires an exponential amount of data to learn a robust classifier. Based on
this theoretical result, we explore how well robust classifiers generalize on
datasets such as CIFAR-10. We come to the conclusion that on this datasets, the
limitation of current robust models also lies in the generalization, and that
they require a lot of data to do well on the test set. We also show that the
problem is not in the expressiveness or generalization capabilities of current
architectures, and that there are low magnitude features in the data which are
useful for non-robust generalization but are not available for robust
classifiers.},
  author       = {Prach, Bernd and Lampert, Christoph},
  booktitle    = {arXiv},
  title        = {{Intriguing properties of robust classification}},
  doi          = {10.48550/arXiv.2412.04245},
  year         = {2024},
}

@inproceedings{18875,
  abstract     = {Current state-of-the-art methods for differentially private model training are based on matrix factorization techniques. However, these methods suffer from high computational overhead because they require numerically solving a demanding optimization problem to determine an approximately optimal factorization prior to the actual model training. In this work, we present a new matrix factorization approach, BSR, which overcomes this computational bottleneck. By exploiting properties of the standard matrix square root, BSR allows to efficiently handle also large-scale problems. For the key scenario of stochastic gradient descent with momentum and weight decay, we even derive analytical expressions for BSR that render the computational overhead negligible. We prove bounds on the approximation quality that hold both in the centralized and in the federated learning setting. Our numerical experiments demonstrate that models trained using BSR perform on par with the best existing methods, while completely avoiding their computational overhead.},
  author       = {Kalinin, Nikita and Lampert, Christoph},
  booktitle    = {38th Annual Conference on Neural Information Processing Systems},
  issn         = {1049-5258},
  location     = {Vancouver, Canada},
  publisher    = {Neural Information Processing Systems Foundation},
  title        = {{Banded square root matrix factorization for differentially private model training}},
  volume       = {37},
  year         = {2024},
}