Generative Modeling
Generative modeling aims to create new data instances that resemble a given dataset, focusing on learning the underlying probability distribution. Current research emphasizes hybrid approaches combining the strengths of autoregressive models (for global context) and diffusion models (for high-quality local details), as well as advancements in flow-based models and score-based methods. These techniques are significantly impacting diverse fields, including image generation, 3D modeling, time series forecasting, and even scientific applications like molecular dynamics simulation and medical image synthesis, by enabling the creation of realistic and diverse synthetic data.
Papers
Latent Diffusion Models for Structural Component Design
Ethan Herron, Jaydeep Rade, Anushrut Jignasu, Baskar Ganapathysubramanian, Aditya Balu, Soumik Sarkar, Adarsh Krishnamurthy
DreamLLM: Synergistic Multimodal Comprehension and Creation
Runpei Dong, Chunrui Han, Yuang Peng, Zekun Qi, Zheng Ge, Jinrong Yang, Liang Zhao, Jianjian Sun, Hongyu Zhou, Haoran Wei, Xiangwen Kong, Xiangyu Zhang, Kaisheng Ma, Li Yi
Score Mismatching for Generative Modeling
Senmao Ye, Fei Liu
A Review of Change of Variable Formulas for Generative Modeling
Ullrich Köthe
Diffusion probabilistic models enhance variational autoencoder for crystal structure generative modeling
Teerachote Pakornchote, Natthaphon Choomphon-anomakhun, Sorrjit Arrerut, Chayanon Atthapak, Sakarn Khamkaeo, Thiparat Chotibut, Thiti Bovornratanaraks