Sparsity in Adaptive Control
Philippe Preux and Sertan Girgin
In: Sparsity in Machine Learning and Statistics workshop, 1-3 Apr, Cumberland Lodge, UK.

Abstract

We investigate methods and algorithms to obtain sparse representations in the context of adaptive control. We are particularly interested in situations in which we look for a control in an unknown, stochastic, possibly non stationary environments, using no prior knowledge. Here, we present our work based on the use of cascade-correlation networks which yields very sparse representations, yet keeping the ability to obtain highly performing controls.

[Edit]