DIRECTIONAL CLASSIFICATION OF CORTICAL SIGNALS USING A LIQUID STATE MACHINE

Abstract

Liquid State Machine (LSM) is a newly developed computational model with many interesting properties. It has great advantages of dealing with biologic computing when compared to the traditional computational model. In this paper, the LSM was used to deal with the direction classification problem of the spike series which were distilled from the neurons in motor cortex of a monkey. In the output layer, a linear regression and back-propagation are employed as the training algorithms. Compare to outcomes of the two algorithms, it is showed that ideal classification results were derived when using BP as the training algorithm.