Barchi Library, 140 John Morgan Building
Center for Computational Biology, Neuroscience Group
Neural computation as transformation of similarity
Making sense of recent large-scale observations of neuronal activity requires a theoretical framework. Shortcomings of the conventional decoding approach motivated us to develop a novel approach based on similarity between stimuli representations. We postulate that neural computation can be viewed as the transformation of similarities. From this postulate, we derive online algorithms that can account for neuroscience observations and are competitive with the state of the art in machine learning.
A pizza lunch will be served.