energy-based models
These probabilistic models define a distribution over data by associating low energy values with high probability regions, providing a framework for learning relationships in data through the minimization of an energy function.
- A Difference-of-Convex Functions Approach to Energy-Based Iterative Reasoning
- AmorLIP: Efficient Language-Image Pretraining via Amortization
- Convex Potential Mirror Langevin Algorithm for Efficient Sampling of Energy-Based Models
- Diffusion Federated Dataset
- Energy Matching: Unifying Flow Matching and Energy-Based Models for Generative Modeling
- Follow the Energy, Find the Path: Riemannian Metrics from Energy-Based Models
- Normalizing Flows are Capable Models for Continuous Control
- Rethinking Entropy in Test-Time Adaptation: The Missing Piece from Energy Duality