Event
Kang-Hao Peng - MS Thesis Defense
Thursday, August 4, 2016
1:30 p.m.
2328 AVW
Melanie Prange
301 405 3686
mprange@umd.edu
Announcement: MS Thesis Defense
Name: Kang-Hao Peng
Committee: Professor Rama Chellappa (chair), Professor Joseph JaJa, Professor Prakash Narayan, Professor Richard La
Date/Time: August 4, 2016, 1:30 pm
Location: 2328 AVW
Title: Channel Capacity-Based RBM Neural Networks
Abstract:
(Deep) neural networks are increasingly being used for various computer vision and pattern recognition tasks due to their strong ability to learn highly discriminative features. However, quantitative analysis of their classification ability and design philosophies is still nebulous. In this work, we use information theory to analyze channel capacity of concatenated restricted Boltzmann machines (RBMs) and propose a channel capacity-based RBM neural networks (CC-RBM). We develop a novel pre-training algorithm to maximize the mutual information between RBMs. Extensive experimental results on various classification tasks show the effectiveness of the proposed approach.