Interval Computing in Neural Networks: One Layer Interval Neural Networks

• Published in: Intelligent Information Technology pp. 68 - 75
• Main Authors: Patiño-Escarcina, Raquel E., Callejas Bedregal, Benjamín R., Lyra, Aarão
• Format: Book Chapter Conference Proceeding
• Language: English
• Published: Berlin, Heidelberg Springer Berlin Heidelberg 01.01.2004; Springer
• Series: Lecture Notes in Computer Science
• Subjects: Applied sciences; Artificial intelligence; Computer science; control theory; systems; Exact sciences and technology; Learning and adaptive systems; Numerical control; Neural network; Supervised learning; Numerical data; Distributed computing; Artificial intelligence
• ISBN: 9783540241263; 3540241264
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-540-30561-3_8

Several applications need a guaranty of the precision of their numerical data. Important tools which allow control of the numerical errors are dealing these data as intervals. This work presents a new approach to use with Interval Computing in Neural Networks, studying the particular case of one layer interval neural networks, which extend Punctual One Layer Neural Networks, and try to be a solution for the problems in calculus precision error and treatment of interval data without modify it. Beyond it, seemly, interval connections between neurons permit the number of the epochs needed to converge to be lower than the needed in punctual networks without loss efficiency. The interval computing in a one layer neural network with supervised training was tested and compared with the traditional one. Experiences show that the behavior of the interval neural network is better than the traditional one beyond of include the guarantee about the computational errors.