Characterizing the Edge of Stability in Variational Training Without Priors
cs.LG, stat.ML
Submitted: 2026-11-11
Updated: 2026-11-11
Code: https://github.com/inferno-ml/inferno
Terminology
Sources
- What Are Bayesian Neural Network Posteriors Really Like?
- On the Expressiveness of Approximate Inference in Bayesian Neural Networks
- Variational Deep Learning via Implicit Regularization
- Why neural networks find simple solutions: the many regularizers of geometric complexity
- The Implicit Bias of Minima Stability in Multivariate Shallow ReLU Networks
- Computing Nonvacuous Generalization Bounds for Deep (Stochastic) Neural Networks with Many More Parameters than Training Data
- On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima
- Sharpness-Aware Minimization for Efficiently Improving Generalization
- The Break-Even Point on Optimization Trajectories of Deep Neural Networks
- The Implicit Regularization of Stochastic Gradient Flow for Least Squares
- Generalization Ability of Wide Neural Networks on $\mathbb{R}$
- On the Origin of Implicit Regularization in Stochastic Gradient Descent
- Adaptive Gradient Methods at the Edge of Stability
- Understanding Optimization in Deep Learning with Central Flows
- On the Relation Between the Sharpest Directions of DNN Loss and the SGD Step Length
- Small Batch Size Training for Language Models: When Vanilla SGD Works, and Why Gradient Accumulation Is Wasteful
- A Tail-Index Analysis of Stochastic Gradient Noise in Deep Neural Networks
- A Simple Baseline for Bayesian Uncertainty in Deep Learning
- Bayesian Deep Learning and a Probabilistic Perspective of Generalization
- Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis