Performance Analysis of Multilayer Neural Net Classifiers using Simulated Pattern-Generating Processes


The Transactions of the Korea Information Processing Society (1994 ~ 2000), Vol. 4, No. 2, pp. 456-464, Feb. 1997
10.3745/KIPSTE.1997.4.2.456,   PDF Download:

Abstract

We describe a random process model that provides sets of patterns with precisely controlled within-class variability and between-class distinctions. We used these patterns in a simulation study with the back-propagation network to characterize its performance as we varied the process-controlling parameters, the statistical differences between the processes, and the random noise on the patterns. Our results indicated that the generalized statistical difference between the processes generating the patterns provided a good predictor of the difficulty of the classification problem. Also we analyzed the performance of the Bayes classifier with the maximum-likelihood criterion and we compared the performance of the neural network to that of the Bayes classifier. We found that the performance of neural was intermediate between that of the simulated and theoretical Bayes classifier.


Statistics
Show / Hide Statistics

Statistics (Cumulative Counts from September 1st, 2017)
Multiple requests among the same browser session are counted as one view.
If you mouse over a chart, the values of data points will be shown.


Cite this article
[IEEE Style]
P. D. Sun, "Performance Analysis of Multilayer Neural Net Classifiers using Simulated Pattern-Generating Processes," The Transactions of the Korea Information Processing Society (1994 ~ 2000), vol. 4, no. 2, pp. 456-464, 1997. DOI: 10.3745/KIPSTE.1997.4.2.456.

[ACM Style]
Park Dong Sun. 1997. Performance Analysis of Multilayer Neural Net Classifiers using Simulated Pattern-Generating Processes. The Transactions of the Korea Information Processing Society (1994 ~ 2000), 4, 2, (1997), 456-464. DOI: 10.3745/KIPSTE.1997.4.2.456.