Transparent Neural Networks, an Implementation

Abstract

The present work is related to a research project denominated Transparent Neural Networks (Strannegård 2012); this project aims to propose a model capable of higher cognitive functions such as deductive and inductive reasoning by means of transparent, simple and interpretable structures and functionalities. This simplicity includes interactive building rules based on the manipulation of basic structures and elements. It considers characteristics of the developmental robotics and cognitive modeling. Taking all of these concepts and goals as a basis, the main objective of this thesis is to help in the development of that model by implementing a toolbox that allows the creation and evaluation of the networks leading to conclusions and meaningful feedback that fosters proposals for further development on the model.