Charles W. Anderson

TL;DR: This work presents the restricted gradient-descent (RGD) algorithm, a training method for local radial-basis function networks specifically developed to be used in the context of reinforcement learning, and shows that the RGD algorithm consistently generates better value-function approximations than the standard gradient- Descent method, and that the latter is more susceptible to divergence.

TL;DR: In this article, the authors train an artificial neural network (ANN) to identify the year of input maps of temperature and precipitation from forced climate model simulations and apply a neural network visualization technique (layerwise relevance propagation) to visualize the spatial patterns that lead the ANN to successfully predict the year.

TL;DR: Experimental results show that up to 90% successful prediction is possible with neural networks and k-nearest neighbor algorithms, improving upon prediction strategies in prior work by approximately 50%.

TL;DR: The stability of a control loop including a recurrent neural network (NN) is analyzed by replacing the nonlinear and time-varying components of the NN with IQCs on their gain and an algorithm is demonstrated for training the recurrent NN using reinforcement learning and guaranteeing stability while learning.

TL;DR: It is demonstrated that learning a predictive model of state dynamics can result in a pretrained hidden layer structure that reduces the time needed to solve reinforcement learning problems.