| Learning in Stochastic Bit Stream Neural Networks. | |
| | |
MedLine Citation:
|
PMID: 12662578 Owner: NLM Status: Publisher |
Abstract/OtherAbstract:
|
This paper presents learning techniques for a novel feedforward stochastic neural network. The model uses stochastic weights and the "bit stream" data representation. It has a clean analysable functionality and is very attractive with its great potential to be implemented in hardware using standard digital VLSI technology. The design allows simulation at three different levels and learning techniques are described for each level. The lowest level corresponds to on-chip learning. Simulation results on three benchmark MONK's problems and handwritten digit recognition with a clean set of 500 16 x 16 pixel digits demonstrate that the new model is powerful enough for the real world applications. Copyright 1996 Elsevier Science Ltd |
| | |
Authors:
|
Max van Daalen; John Shawe-Taylor; Jieyu Zhao |
Related Documents
:
|
465578 - Heart rate and blood pressure biofeedback: ii. a review and integration of recent theor... 12136998 - Individuality of handwriting. 20351798 - A framework for learning from distributed data using sufficient statistics and its appl... 19474048 - Song learning in domesticated canaries in a restricted acoustic environment. 23481208 - Nuclear organization in the nematode c. elegans. 9569938 - Generalization of independent response model for toxic mixtures. |
Publication Detail:
|
Type: JOURNAL ARTICLE |
Journal Detail:
|
Title: Neural networks : the official journal of the International Neural Network Society Volume: 9 ISSN: 1879-2782 ISO Abbreviation: Neural Netw Publication Date: 1996 Aug |
Date Detail:
|
Created Date: 2003-Mar-28 Completed Date: - Revised Date: - |
Medline Journal Info:
|
Nlm Unique ID: 8805018 Medline TA: Neural Netw Country: - |
Other Details:
|
Languages: ENG Pagination: 991-998 Citation Subset: - |
Affiliation:
|
University of London, Egham, Surrey, TW20 0EX, UK |
Export Citation:
|
APA/MLA Format Download EndNote Download BibTex |
| MeSH Terms | |
Descriptor/Qualifier:
|
|
From MEDLINE®/PubMed®, a database of the U.S. National Library of Medicine
Previous Document: A New Family of Multivalued Networks.
Next Document: A Decomposition Principle for Complexity Reduction of Artificial Neural Networks.