The Accuracy Well: Nonmonotone Accuracy Dynamics at Vanishing Initialization

arXiv:2610.20439 · cs.LG · Submitted 2026-10-14 · Read on arXiv

cs.LG

Submitted: 2026-10-14

Updated: 2026-10-14

Comments: 124 pages, 11 figures. Submitted to ICLR-2027: https://openreview.net/forum?id=OZ6JYRP75V

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Accuracy — 2× in this paper · explained in 1 episode

Sources

• Implicit Regularization of Discrete Gradient Dynamics in Linear Neural Networks
• Saddle-to-Saddle Dynamics in Deep Linear Networks: Small Initialization Training, Symmetry, and Sparsity
• Gradient Descent Quantizes ReLU Network Features
• Towards Understanding Epoch-wise Double descent in Two-layer Linear Neural Networks
• Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets
• An Empirical Study of Example Forgetting during Deep Neural Network Learning

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks