In this course we will examine how biological neural circuits and brain function can affect the design of machine learning hardware and software, and vice versa. We will attempt to build a better understanding of how similar and different the computational approaches of the two are, and what can be deduced from one area about the other. Examples will be biological constructs such as biological neurons, cortical columns, connectomes, associative memory, and natural processes like pruning, versus artificial neural network hardware and software designs. We will look at plausible alternative learning mechanisms to backpropagation such as feedback alignment, equilibrium propagation and predictive coding, and try to better understand state of the art machine learning optimization techniques such as sparsification and quantization.

The class will be in a seminar style, based on a select collection of papers on the topic.