Skip to main content

Multiscale Adaptive Representation of Signals: I. The Basic Framework

Author(s): Tai, Cheng; E, Weinan

Download
To refer to this page use: http://arks.princeton.edu/ark:/88435/pr1ks91
Full metadata record
DC FieldValueLanguage
dc.contributor.authorTai, Cheng-
dc.contributor.authorE, Weinan-
dc.date.accessioned2017-11-21T19:41:40Z-
dc.date.available2017-11-21T19:41:40Z-
dc.date.issued2016-01en_US
dc.identifier.citationTai, Cheng, E, Weinan. (2016). Multiscale Adaptive Representation of Signals: I. The Basic Framework. JOURNAL OF MACHINE LEARNING RESEARCH, 17en_US
dc.identifier.issn1532-4435-
dc.identifier.urihttp://arks.princeton.edu/ark:/88435/pr1ks91-
dc.description.abstractWe introduce a framework for designing multi-scale, adaptive, shift-invariant frames and bi-frames for representing signals. The new framework, called AdaFrame, improves over dictionary learning-based techniques in terms of computational efficiency at inference time. It improves classical multi-scale basis such as wavelet frames in terms of coding efficiency. It provides an attractive alternative to dictionary learning-based techniques for low level signal processing tasks, such as compression and denoising, as well as high level tasks, such as feature extraction for object recognition. Connections with deep convolutional networks are also discussed. In particular, the proposed framework reveals a drawback in the commonly used approach for visualizing the activations of the intermediate layers in convolutional networks, and suggests a natural alternative.en_US
dc.format.extent4875-4912en_US
dc.language.isoenen_US
dc.relation.ispartofJOURNAL OF MACHINE LEARNING RESEARCHen_US
dc.rightsAuthor's manuscripten_US
dc.titleMultiscale Adaptive Representation of Signals: I. The Basic Frameworken_US
dc.typeJournal Articleen_US
dc.date.eissued2016-08en_US
pu.type.symplectichttp://www.symplectic.co.uk/publications/atom-terms/1.0/journal-articleen_US

Files in This Item:
File Description SizeFormat 
1507.04835v1.pdf1.51 MBAdobe PDFView/Download


Items in OAR@Princeton are protected by copyright, with all rights reserved, unless otherwise indicated.