Neuroscience Group Meeting
Presenter: David Lipshutz
Title: Journal club presentation: Neural network learning of optimal Kalman prediction and control
Description: We discuss the paper “Neural network learning of optimal Kalman prediction and control” by Ralph Pinsker (2008). Kalman prediction and control are classical solutions for optimal estimation and control of linear systems with noisy measurements. In this paper, the author proposes an algorithm for a recurrent neural network to learn and implement Kalman prediction and control. The structure of the neural network for implementing Kalman control bears certain resemblances to the local cortical circuit, which the author explores.