Radial Basis Neural Network for Lossless Data Compression

• Published in: International journal of computers & applications 2002-01, Vol.24 (1), p.14-19
• Main Authors: Logeswaran, R., Eswaran, C.
• Format: Article
• Language: English
• Subjects: Applied sciences; Artificial intelligence; Computer science; control theory; systems; Connectionism. Neural networks; data compression; Exact sciences and technology; Information, signal and communications theory; lossless compression; Miscellaneous; Neural networks; predictor; radial basis network; Signal processing; Telecommunications and information theory; two-stage compression
• ISSN: 1206-212X; 1925-7074
• DOI: 10.1080/1206212X.2002.11441655

The radial basis network is essentially a function approximator; article shows that this characteristic can be exploited for data compression applications. A variant of the radial basis network, the generalized regression neural network, is used in a two-stage compression scheme and its performance is evaluated in terms of the compression ratio. The training is imparted to the network using a block adaptive training method, and the trained network performs as a predictor-approximator in the first stage of compression. The optimum configuration of the network is arrived at by using a trial-and-error procedure. The compression ratios achieved by this network when used along with an arithmetic encoder in a two-stage compression scheme are obtained for different test files containing telemetry data. It is found that these results are comparable to those obtained with other known classical linear predictors.