Mamba in a nutshell
Organizer
Speaker
(Tsinghua) Tao Zhou
Time
Monday, May 20, 2024 2:30 PM - 3:00 PM
Venue
理科楼A-304
Abstract
We will introduce Mamba, a sequential language model that was proposed at the end of 2023. The model is based on the classical state-space model in cybernetics, complemented by fine-grained mathematical tools that achieve model performance that rivals that of transformers in terms of linear time complexity. In the original paper, the authors also showed the potential of Mamba for modeling DNA sequences, and just two weeks ago, a Mamba-based DNA sequence model, Caduceus, was also proposed. Therefore, it is necessary to learn the ideas embedded in Mamba and explore the possibility of applying Mamba models to more diverse bioinformatics problems in the future.