The research of the Neural Information Processing group focuses on two aspects: one is to use mathematical models to elucidate the mechanisms of brain functions, and the other is to develop brain-style computational algorithms. We will collaborate closely with the experimental groups in ION and in other places to identify interesting and important questions in biological computation, and to use or develop mathematical models to solve them.
Currently, we are working on three projects, but more are likely to be added in the future. These three projects are:
To develop a network theory that describes the encoding of objects and their categorization in neural systems. In particular, we will study the mathematical properties of continuous attractor neural networks and their applications in neural information processing. We will also carry out psychophysical experiments to confirm the theoretical predictions.
To develop advanced machine learning methods for extracting information from neural-imaging data. In particular, we will develop methods to improve the performance of Support Vector Machine.
To develop a network theory that elucidates how information is dynamically routed in neural systems in order to achieve optimal computation. In particular, we will study how dynamical routing implements the compensation of the transmission delay in neural signals.