Mutual Information-based RBM Neural Networks

dc.contributor.advisorChellappa, Ramaen_US
dc.contributor.authorPeng, Kang-Haoen_US
dc.contributor.departmentElectrical Engineeringen_US
dc.contributor.publisherDigital Repository at the University of Marylanden_US
dc.contributor.publisherUniversity of Maryland (College Park, Md.)en_US
dc.date.accessioned2016-09-15T05:35:08Z
dc.date.available2016-09-15T05:35:08Z
dc.date.issued2016en_US
dc.description.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 classication ability and design philosophies are still nebulous. In this work, we use information theory to analyze the concatenated restricted Boltzmann machines (RBMs) and propose a mutual information-based RBM neural networks (MI-RBM). We develop a novel pretraining algorithm to maximize the mutual information between RBMs. Extensive experimental results on various classication tasks show the eectiveness of the proposed approach.en_US
dc.identifierhttps://doi.org/10.13016/M25V4B
dc.identifier.urihttp://hdl.handle.net/1903/18838
dc.language.isoenen_US
dc.subject.pqcontrolledElectrical engineeringen_US
dc.subject.pqcontrolledComputer scienceen_US
dc.subject.pquncontrolledDeep Learningen_US
dc.subject.pquncontrolledMutual Informationen_US
dc.subject.pquncontrolledNeural Networken_US
dc.subject.pquncontrolledRestricted Boltzmann Machineen_US
dc.titleMutual Information-based RBM Neural Networksen_US
dc.typeThesisen_US

Files

Original bundle

Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
Peng_umd_0117N_17574.pdf
Size:
281.51 KB
Format:
Adobe Portable Document Format