Optimal Condition for Initialization Variance in Deep Neural Networks: An SGD Dynamics Perspective

arXiv:2508.12834 · stat.ML, cs.LG · Submitted 2025-08-18 · Read on arXiv

stat.ML, cs.LG

Submitted: 2025-08-18

Updated: 2026-10-14

Terminology

• Accuracy — 1× in this paper · explained in 1 episode
• Kullback-Leibler (KL) Divergence — 1× in this paper · explained in 4 episodes
• Langevin Dynamics — 1× in this paper · explained in 2 episodes
• Loss Function — 1× in this paper · explained in 2 episodes
• Stochastic Gradient Descent (SGD) — 1× in this paper · explained in 5 episodes

Related papers

• Behavior of prediction performance metrics with rare events
• Optimal Estimation of Generic Dynamics by Path-Dependent Neural Jump ODEs
• A Posterior-Dynamics Framework for Imaging Inverse Problems with Pretrained Diffusion Priors
• Efficient quadratic entropy with distance sketches
• Embedding-Bias in Conditional Independence Testing
• From Log-Odds to Shapley Values: An Explanatory Geometry for the Weighted Naive Bayes Classifier