Showing papers by "Charles W. Anderson published in 2007"

Classification of EEG Signals from Four Subjects During Five Mental Tasks

Abstract: Neural networks are trained to classify half-second segments of six-channel, EEG data into one of five classes corresponding to five cognitive tasks performed by four subjects. Two and three-layer feedforward neural networks are trained using 10-fold cross-validation and early stopping to control over-fitting. EEG signals were represented as autoregressive (AR) models. The average percentage of test segments correctly classified ranged from 71% for one subject to 38% for another subject. Cluster analysis of the resulting neural networks’ hidden-unit weight vectors identifies which EEG channels are most relevant to this discrimination problem.

Robust Reinforcement Learning Control Using Integral Quadratic Constraints for Recurrent Neural Networks

Abstract: The applicability of machine learning techniques for feedback control systems is limited by a lack of stability guarantees. Robust control theory offers a framework for analyzing the stability of feedback control loops, but for the integral quadratic constraint (IQC) framework used here, all components are required to be represented as linear, time-invariant systems plus uncertainties with, for IQCs used here, bounded gain. In this paper, 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. As a result, a range of the NN's weights is found within which stability is guaranteed. An algorithm is demonstrated for training the recurrent NN using reinforcement learning and guaranteeing stability while learning.

Improving performance using robust recurrent reinforcement learning control

Abstract: A recurrent neural network (RNN) is used inside the feedback loop to improve closed-loop tracking performance of a nonlinear plant. An actor-critic reinforcement learning algorithm is used to optimize the RNN actor as the plant operates in real-time. Integral Quadratic Constraints (IQCs) are used to guarantee robust stability of the closed-loop system as the RNN actor learns online. Using IQCs, we can deal with both model uncertainty and the nonlinear elements of the RNN actor in a single unified framework. The RNN actor provides dynamic capabilities that a feed-forward neural network could not, which results in a robust recurrent reinforcement learning controller that is able to provide enhanced nonlinear dynamic control.