Research
What is compositionality? It is a kind of representation format. Based on compositional coding, neural representation could be decomposed into different compartments, and each compartment keep relatively independent. In this case, when one compartment change, others keep still. Compositional coding will help our brains to process sequence information, and also help our brains to do transfer learning.
Research interest 1: Sequences are the organized by their ordinal structure and contents. Like human language, these sequences consist of syntaxes (ordinal structures) and words (contents). With compositionality, limited words and syntaxes could create infinite language. One of our goals is to understand the neural mechanisms when primates learn and memorize a bunch of sequences with simple or complex ordinal structures.
Research interest 2: Our brain could learn multiple tasks with limited samples in short time. These tasks could be similar or different, could be basic or sophisticated. How neural populations engaged in multiple tasks? Is one ensemble responsible for one task, or same multiple ensembles compositionally deal with multiple tasks? We will try to train animals to deal with multi-task problem, and probe underlying neural mechanism.
Methodologies: We will try to probe the neural basis of compositionality with large scale recording. With two-photon imaging or large scale electrophysiological recording, we could record hundreds of neurons simultaneously.