Is this you? As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.
Claim your profile
Articles by Daniel Soudry
Computer Science > Machine Learning
We study continual linear regression in a high-dimensional regime, where a learner observes a sequence of tasks and updates its parameters sequentially. In this setting, we derive an explicit closed-form expression for the generalization loss of a model trained with gradient descent for a fixed number of steps per task, which holds for arbitrary teachers.
Computer Science > Machine Learning
We characterize weakly regularized continual classification in homogeneous models as sequential projections onto task margin sets. This result generalizes prior analyses restricted to either stationary (single-task) deep models or continual linear models. We show that global convergence generally fails, even for simple models linear in data but nonlinear in parameters.
Computer Science > Machine Learning
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Neuer Gesetzesentwurf zur Beschleunigung von Bundeswehr-Vergaben
Am 23.07.2025 hat die Bundesregierung den Referentenentwurf für ein weiteres Gesetz zur Beschleunigung von Vergabeverfahren der Bundeswehr beschlossen. Der gemeinsame Entwurf des BMWE und des BMVg vom 26.06.2025 übernimmt wesentliche Teile des zeitlich befristeten Vorgängergesetzes auf Dauer und geht in Teilen darüber hinaus. Die Rechte von Unternehmen werden erneut empfindlich verkürzt.
[2505.19087] Temperature is All You Need for Generalization in Langevin Dynamics and other Markov Processes
Work on one of the world's most important websites and make an impact on open science.
[2409.12517] Scaling FP8 training to trillion-token LLMs
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
[2401.12617] The Joint Effect of Task Similarity and Overparameterization on Catastrophic Forgetting -- An Analytical Model
arXiv:2401.12617 (cs) [Submitted on 23 Jan 2024 (v1), last revised 24 Jan 2024 (this version, v2)] Download PDF Comments: Subjects: Machine Learning (cs.LG) Cite as: arXiv:2401.12617 [cs.LG] (or arXiv:2401.12617v2 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2401.12617 Submission history From: Itay Evron [ view email] [v1] Tue, 23 Jan 2024 10:16:44 UTC (2,266 KB) [v2] Wed, 24 Jan 2024 12:49:24 UTC (2,266 KB) Bibliographic Tools Bibliographic Explorer Toggle Bibliographic Explorer...
Ein Gegner wie die Hamas bringt das humanitäre Völkerrecht an seine Grenze
Microsoft tritt für den Schutz Ihrer Privatsphäre ein. Microsoft und unsere Drittanbieter verwenden Cookies, um Informationen wie eindeutige IDs zu speichern und darauf zuzugreifen, um unsere Dienste und Anzeigen bereitzustellen, zu pflegen und zu verbessern. Wenn Sie zustimmen, personalisieren MSN und Microsoft Bing die angezeigten Inhalte und Anzeigen.
Krieg zwischen Israel und der Hamas: Stößt das Humanitäre Völkerrecht an seine Grenzen?
Der Großangriff auf Israel am 7. Oktober ist von beispiellosem Ausmaß. Mehr als 1200 Zivilisten und über 300 Soldaten starben, als die Hamas Kibbuz-Gemeinden und Städte in der Nachbarschaft des Gazastreifens überfiel. Die Terroristen mordeten nicht bloß: Sie erniedrigten ihre Opfer unbarmherzig und sadistisch durch Folter, Verstümmelung und sexualisierte Gewalt.
[1912.05137] Is Feature Diversity Necessary in Neural Network Initialization?
[Submitted on 11 Dec 2019 (v1), last revised 12 Dec 2019 (this version, v2)] Authors:Yaniv Blumenfeld, Dar Gilboa, Daniel Soudry Download PDF Abstract: Standard practice in training neural networks involves initializing the weights in an independent fashion. The results of recent work suggest that feature "diversity" at initialization plays an important role in training the network. However, other initialization schemes with reduced feature diversity have also been shown to be viable.
Simultaneous Denoising, Deconvolution, and Demixing of Calcium Imaging Data
The proposed AR framework makes a number of simplifying assumptions on the fluorescence dynamics, with the benefit of increased computational tractability. The dynamics are assumed to be linear and time invariant, and no saturation level is assumed.
Binarized Neural Networks: Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1
arXiv:1602.02830 (cs) [Submitted on 9 Feb 2016 (v1), last revised 17 Mar 2016 (this version, v3)] Download PDF Comments: Subjects: Machine Learning (cs.LG) Cite as: arXiv:1602.02830 [cs.LG] (or arXiv:1602.02830v3 [cs.LG] for this version) https://doi.org/10.48550/arXiv.1602.02830 Submission history From: Itay Hubara [ view email] [v1] Tue, 9 Feb 2016 01:01:59 UTC (71 KB) [v2] Mon, 29 Feb 2016 21:26:53 UTC (94 KB) [v3] Thu, 17 Mar 2016 14:54:25 UTC (94 KB) Bibliographic Tools Bibliographic...
Show More
loading
Actions
Is this you?
As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.Get in touch with Daniel
Contact Daniel, search articles and posts on X, monitor coverage, and track replies from one place.
Learn more about Muck Rack