Combining ESOMs Trained on a Hierarchy of Feature Subsets for Single-Trial Decoding of LFP Responses in Monkey Area V4

Abstract

We develop and combine topographic maps trained on different combinations of feature subsets for visualizing and classifying event-related responses recorded with a multi-electrode array chronically implanted in the visual cortical area V4 of a rhesus monkey. The monkey was trained, during consecutive training sessions, in a classical conditioning paradigm in which one stimulus was consistently paired with a fluid reward and another stimulus not. We opted for features from three categories: time-frequency analysis, phase synchronization between electrodes, and propagating waves in the array. The Emergent Self Organizing Map (ESOM) was used to explore the feasibility of single-trial decoding. Since the effective dimensionality of the feature space is rather high, a series of ESOMs was trained on features selected from different combinations of the three feature categories. For each trained ESOM, a classifier was developed, and classifiers of different ESOMs were combined so as to maximize the single-trial decoding performance.