Interval Computing in Neural Networks: One Layer Interval Neural Networks

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