|
Abstract: |
| In this talk, I will review my exploration experience in biomolecular design. Through the transformation of the core modeling paradigm, significant gains can often be achieved. For instance, non-autoregressive decoding in protein inverse folding, K-Rook models in RNA secondary structure prediction, and cross-attention in electron microscope structure analysis, etc. Finally, I will share my thoughts on four important research directions in the field of biomolecular design, ranging from the acquisition of experimental data to the prediction of structural systemic bias debiasing. From post-training refinement of concrete structures to abstract structure design. These four directions together form the vision from the data foundation to advanced design capabilities. |
Biography： |
Dr. Cheng Tan is a research scientist at Shanghai AI Lab, holding a joint Ph.D. from Zhejiang University and Westlake University under Professor Stan Z. Li. He has published over ten papers as first or co-first author in top journals such as TPAMI and Nature sub-journals, with more than 3,500 citations and an H-index of 25. His work has garnered multiple top conference presentations, and he has received the National Scholarship from Zhejiang University. Dr. Tan has led a project funded by the National Natural Science Foundation of China and serves as a reviewer for prestigious conferences and journals like NeurIPS, ICML, and CVPR. |