deep generative models
Deep generative models are a class of models designed to learn to generate new data points that mimic a given dataset. Techniques include Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs), which have applications in creating anything from images to textual data.
- CDFlow: Building Invertible Layers with Circulant and Diagonal Matrices
- Cue3D: Quantifying the Role of Image Cues in Single-Image 3D Generation
- Flexible MOF Generation with Torsion-Aware Flow Matching
- Learning conformational ensembles of proteins based on backbone geometry
- On the Closed-Form of Flow Matching: Generalization Does Not Arise from Target Stochasticity
- On the Closed-Form of Flow Matching: Generalization Does Not Arise from Target Stochasticity
- Physics-Constrained Flow Matching: Sampling Generative Models with Hard Constraints
- Principled Long-Tailed Generative Modeling via Diffusion Models
- Protein Design with Dynamic Protein Vocabulary
- Simultaneous Modeling of Protein Conformation and Dynamics via Autoregression