A CONTEXT-INTEGRATING SIGNAL CLASSIFICATION MODEL FOR RESOLVING AMBIGUOUS STIMULI

Abstract

The brain uses contextual information to uniquely resolve the interpretation of ambiguous stimuli. An interdisciplinary effort which combines expertise in machine learning and neuroscience is used to formulate a generalized signal classification model that has the ability to integrate weighted bidirectional temporal or spatial context to effectively resolve the classification of ambiguous stimuli. The formulation of the model is quite general; consequently, it is not restricted to stimuli in any particular sensory modality nor to any type of classifier. Furthermore, the model parameters can be manipulated to simulate various context environments. The context-integrating model is implemented using a Gaussian multivariate classifier and a broad range of experiments are designed to demonstrate its effectiveness in classifying ambiguous visual stimuli in various contextual environments.