2021

(Anonymous Paper)

Shitong Luo Jiaqi Guan Jianzhu Ma Jian Peng

Submitted to NeurIPS 2021

Drug design.

(Anonymous Paper)

Shitong Luo* Chence Shi* Minkai Xu Jian Tang (* indicates equal contribution)

Submitted to NeurIPS 2021

Physically based generative modeling.

EBM-Fold: Fully-Differentiable Protein Folding Powered by Energy-based Models

Jiaxiang Wu Shitong Luo Tao Shen Haidong Lan Sheng Wang Junzhou Huang

arXiv Preprint: 2105.04771 [cs.LG] 2021

Applying our gradient field method (Shi*, Luo*, et al., ICML'21) to protein folding.

Score-Based Point Cloud Denoising

Shitong Luo Wei Hu

International Conference on Computer Vision (ICCV) 2021

Estimating the score of noisy point distributions for effective and robust point cloud denoising and beyond.

Learning Gradient Fields for Molecular Conformation Generation

Chence Shi* Shitong Luo* Minkai Xu Jian Tang (* indicates equal contribution)

International Conference on Machine Learning (ICML) 2021 Oral / Long Talk

End-to-end conformation sampling; Equivariant score matching.

Diffusion Probabilistic Models for 3D Point Cloud Generation

Shitong Luo Wei Hu

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2021 Oral Best Paper Candidate

Inspired by the diffusion process in non-equilibrium thermodynamics; Markov chain; Competitive performance in generation and auto-encoding.

Learning Neural Generative Dynamics for Molecular Conformation Generation

Minkai Xu* Shitong Luo* Yoshua Bengio Jian Peng Jian Tang (* indicates equal contribution)

International Conference on Learning Representations (ICLR) 2021

Flow-based model for molecular conformation generation; Generating inter-atomic distances to preserve roto-translational invariance; High diversity and fidelity.

2020

Differentiable Manifold Reconstruction for Point Cloud Denoising

Shitong Luo Wei Hu

ACM International Conference on Multimedia (ACM MM) 2020 Oral

Neural network architecture for point cloud denoising; Achieving state-of-the-art-performance; Surface-fitting based; Downsample-upsample scheme; Robust against high noise.