MIL - The University of Tennessee
the university of tennessee machine intelligence lab

Deep Learning Architectures

Background and Motivation

Information representation is at the core of many machine learning applications. As such, deep learning networks offer the premise of powerful information representation, in the context of both time and space, so as to empower new machine learning engines in a way that was not previously possible.

Research Approach

Our general approach pertains to a scalable, hierarchical Bayesian-inference deep learning architecture which comprises of a basic cortical circuit that populates the homogeneously hierarchy. Representation of high-dimensional patterns is achieved in a manner driven by regularities in the observations, such that complex features are formed across the hierarchy. Each node in our architecture makes sense of current belief information from the layer above it, in conjunction with the observation seen below. Imperfect observations are recoverable through an online unsupervised learning engine. The results reflect robust pattern recognition capabilities suitable to a wide range of applications.

For a reference on deep learning architectures please see: I. Arel, D. Rose, T. Karnowski, "Deep Machine Learning - A New Frontier in Artificial Intelligence Research," IEEE Computational Intelligence Magazine, Vol. 14, pp. 12-18, November, 2010 [pdf]