A Novel Method for Training an Echo State Network with Feedback-Error Learning

[摘要] Echo state networks are a relatively new type of recurrent neural networksthat have shown great potentials for solving non-linear, temporalproblems. The basic idea is to transform the low dimensional temporal inputinto a higher dimensional state, and then train the output connectionweights to make the system output the target information. Because onlythe output weights are altered, training is typically quick and computationallyefficient compared to training of other recurrent neural networks. This paper investigates using an echo state network to learn the inversekinematics model of a robot simulator with feedback-error-learning. Inthis scheme teacher forcing is not perfect, and joint constraints on thesimulator makes the feedback error inaccurate. A novel training methodwhich is less influenced by the noise in the training data is proposed andcompared to the traditional ESN training method.

[学科分类] 人工智能