2015 Data Science Workshop


Wed, April 29, 2015
Location: Fisher Conference Center, Arrillaga Alumni Center

"Real-Time Large-Scale Neural Identification"


Large­-scale neural recording is transforming how we study and understand the brain and nervous system. With the advent of high­-density electrode arrays, high­-density readout electronics, and the computational ability to analyze and mine large data sets, we have the possibility to understand for the first time the function of entire neural circuits composed of hundreds or thousands of neurons. However, a major technical barrier remains: each electrode typically records from multiple neurons, and each neuron is typically recorded by multiple electrodes. To understand the function of a neural circuit, we must separate the discrete nerve impulses (spikes) produced by distinct cells, a challenging mathematical inverse problem known as “spike sorting”. Previous approaches to spike sorting are approximate, intrinsically error­-prone, and involve laborious manual supervision. Thus, they do not scale to large data sets or real­-time analysis. This project aims to push forward the feasibility boundary of spike sorting techniques for large­-scale neural recordings using modern mathematical and computational methods. We will develop a system that reconstructs and analyzes the activity of a large population of retinal neurons in real time, based on recordings from hundreds or thousands of electrodes. This will guide the selection of new stimuli in closed loop, to probe the specific functions of a neural population and to elucidate the visual processing functions of the retina.