loss landscape
The loss landscape is a visualization of how the loss function varies with different model parameters, impacting optimization by indicating regions of stability or difficulty in finding optimal solutions.
- A Unified Stability Analysis of SAM vs SGD: Role of Data Coherence and Emergence of Simplicity Bias
- Any-stepsize Gradient Descent for Separable Data under Fenchel–Young Losses
- Continual Optimization with Symmetry Teleportation for Multi-Task Learning
- Exploring Landscapes for Better Minima along Valleys
- Flat Channels to Infinity in Neural Loss Landscapes
- Flatness is Necessary, Neural Collapse is Not: Rethinking Generalization via Grokking
- How to Learn a Star: Binary Classification with Starshaped Polyhedral Sets
- Modality-Aware SAM: Sharpness-Aware-Minimization Driven Gradient Modulation for Harmonized Multimodal Learning
- Non-Singularity of the Gradient Descent Map for Neural Networks with Piecewise Analytic Activations
- Not All Data are Good Labels: On the Self-supervised Labeling for Time Series Forecasting
- Optimality and NP-Hardness of Transformers in Learning Markovian Dynamical Functions
- RODS: Robust Optimization Inspired Diffusion Sampling for Detecting and Reducing Hallucination in Generative Models
- RefLoRA: Refactored Low-Rank Adaptation for Efficient Fine-Tuning of Large Models
- RoMa: A Robust Model Watermarking Scheme for Protecting IP in Diffusion Models
- SharpZO: Hybrid Sharpness-Aware Vision Language Model Prompt Tuning via Forward-Only Passes
- Targeted Maximum Likelihood Learning: An Optimization Perspective
- Theoretical Insights into In-context Learning with Unlabeled Data
- Through the River: Understanding the Benefit of Schedule-Free Methods for Language Model Training
- Towards Generalizable 3D Human Pose Estimation via Ensembles on Flat Loss Landscapes
- Understanding and Improving Fast Adversarial Training against $l_0$ Bounded Perturbations