[2503.22478] Almost Bayesian: The Fractal Dynamics of Stochastic Gradient Descent

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. arXiv Operational Status Get status notifications via email or slack

[2503.22478] Almost Bayesian: The Fractal Dynamics of Stochastic Gradient Descent Skip to main content arXiv is now an independent nonprofit! Learn more × Search arXiv Press Enter to search · Advanced search --> Computer Science > Machine Learning arXiv:2503.22478 (cs) [Submitted on 28 Mar 2025 ( v1 ), last revised 16 Mar 2026 (this version, v2)] Title: Almost Bayesian: The Fractal Dynamics of Stochastic Gradient Descent Authors: Max Hennick , Stijn De Baerdemacker View a PDF of the paper titled Almost Bayesian: The Fractal Dynamics of Stochastic Gradient Descent, by Max Hennick a

related reading

• [2404.19756] KAN: Kolmogorov-Arnold Networksarxiv.org
• arxiv.org/pdf/1805.08522arxiv.org
• [2503.11615v2] From Score Matching to Diffusion: A Fine-Grained Error Analysis in the Gaussian Settingarxiv.org
• [2512.19428] Attention Is Not What You Needarxiv.org
• [2508.12773] Online Ensemble Transformer for Accurate Cloud Workload Forecasting in Predictive Auto-Scalingarxiv.org
• arxiv.org/pdf/2512.24880#page=3.56arxiv.org
• [2506.03719] On the Closed-Form of Flow Matching: Generalization Does Not Arise from Target Stochasticityarxiv.org
• [2207.08799] Hidden Progress in Deep Learning: SGD Learns Parities Near the Computational Limitarxiv.org
• [2505.12387] Neural Thermodynamics: Entropic Forces in Deep and Universal Representation Learningarxiv.org
• [2504.18208v1] Ultra-fast feature learning for the training of two-layer neural networks in the two-timescale regimearxiv.org
• [2509.14786] Pre-training under infinite computearxiv.org
• [2604.06366] Stochastic Gradient Descent in the Saddle-to-Saddle Regime of Deep Linear Networksarxiv.org