Searching for just a few words should be enough to get started. If you need to make more complex queries, use the tips below to guide you.
Article type: Research Article
Authors: Ray, Sylvian R.a; * | Hsu, William H.b; 1
Affiliations: [a] Department of Computer Science, University of Illinois at Urbana-Champaign, 1304 West Springfield Avenue, Urbana, IL 61801, USA | [b] National Computational Science Alliance, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA
Correspondence: [*] Corresponding author. E-mail: ray@cs.uiuc.edu.
Note: [1] E-mail: bhsu@ncsa.uiuc.edu.
Abstract: In this paper, we investigate a form of modular neural network for classification with (a) pre-separated input vectors entering its specialist (expert) networks, (b) specialist networks which are self-organized (radial-basis function or self-targeted feedforward type) and (c) which fuses (or integrates) the specialists with a single-layer net. When the modular architecture is applied to spatiotemporal sequences, the Specialist Nets are recurrent; specifically, we use the Input Recurrent type. The Specialist Networks (SNs) learn to divide their input space into a number of equivalence classes defined by self-organized clustering and learning using the statistical properties of the input domain. Once the specialists have settled in their training, the Fusion Network is trained by any supervised method to map to the semantic classes. We discuss the fact that this architecture and its training is quite distinct from the hierarchical mixture of experts (HME) type as well as from stacked generalization. Because the equivalence classes to which the SNs map the input vectors are determined by the natural clustering of the input data, the SNs learn rapidly and accurately. The fusion network also trains rapidly by reason of its simplicity. We argue, on theoretical grounds, that the accuracy of the system should be positively correlated to the product of the number of equivalence classes for all of the SNs. This network was applied, as an empirical test case, to the classification of melodies presented as direct audio events (temporal sequences) played by a human and subject, therefore, to biological variations. The audio input was divided into two modes: (a) frequency (or pitch) variation and (b) rhythm, both as functions of time. The results and observations show the technique to be very robust and support the theoretical deductions concerning accuracy.
Keywords: Modular neural networks, Temporal sequences, Fusion, Multichannel signals
DOI: 10.3233/IDA-1998-2403
Journal: Intelligent Data Analysis, vol. 2, no. 4, pp. 287-301, 1998
IOS Press, Inc.
6751 Tepper Drive
Clifton, VA 20124
USA
Tel: +1 703 830 6300
Fax: +1 703 830 2300
sales@iospress.com
For editorial issues, like the status of your submitted paper or proposals, write to editorial@iospress.nl
IOS Press
Nieuwe Hemweg 6B
1013 BG Amsterdam
The Netherlands
Tel: +31 20 688 3355
Fax: +31 20 687 0091
info@iospress.nl
For editorial issues, permissions, book requests, submissions and proceedings, contact the Amsterdam office info@iospress.nl
Inspirees International (China Office)
Ciyunsi Beili 207(CapitaLand), Bld 1, 7-901
100025, Beijing
China
Free service line: 400 661 8717
Fax: +86 10 8446 7947
china@iospress.cn
For editorial issues, like the status of your submitted paper or proposals, write to editorial@iospress.nl
如果您在出版方面需要帮助或有任何建, 件至: editorial@iospress.nl