A system extending backpropagation to learn classification/categorization without a teacher, form a small basis for larger knowledge structures, and make additional reasoning using knowledge learned from classification/categorization

Summary form only given. The author briefly reviews backpropagation and points out in detail the problems and issues related to it. Then he examines a method that will help alleviate some of these problems, all within the same system. Although the system is explained part by part, the small contribution of this study is perhaps not how each specific part solves a problem, but how the various pieces of the resulting system fit together to alleviate some of those problems within the same system. In addition, the author examines some of the issues related to the resulting method.<>