closed-form solutions
These are analytical solutions expressed as explicit formulas that can be computed in a finite number of standard operations, which are desirable in optimization problems as they facilitate easier interpretation and faster computation.
- An Analytical Theory of Spectral Bias in the Learning Dynamics of Diffusion Models
- Approximate Gradient Coding for Distributed Learning with Heterogeneous Stragglers
- Riemannian Consistency Model
- Robust Reinforcement Learning in Finance: Modeling Market Impact with Elliptic Uncertainty Sets
- Tackling Biased Evaluators in Dueling Bandits
- Theoretically Grounded Framework for LLM Watermarking: A Distribution-Adaptive Approach
- Towards Minimizing Feature Drift in Model Merging: Layer-wise Task Vector Fusion for Adaptive Knowledge Integration
- Tree-Sliced Entropy Partial Transport