AI methods, and in particular deep learning models, have achieved remarkable success in computer vision, speech recognition, and natural language processing due to the availability of powerful computational resources. Recently, neuroscience has entered an exciting new age. Modern recording technologies in neuroscience, like Multi-Electrode Arrays(MEA), enable simultaneous measurements of thousands of neurons’ activities. Such recordings offer an unprecedented opportunity to learn the mechanistic activities of neuron intelligence, but they also present an extraordinary computational and statistical challenge: How do we make sense of these large scale recordings?
In this talk, we will review existing works on using AI methods to interpret neuroscience recording data. In addition, we will review state-of-the-art statistical Bayesian methods for data analysis of multi-neuron recordings. Finally, we will present a preliminary deep learning framework for MEA classification on mouse and human derived induced Pluripotent Stem Cell recordings.