T-DTS (Tree-like Divide to Simplify) is a next software version 3.0 writen in Matlab 6 in the scope of my thesis: "Contribution to the Study and Implementation of Intelligent Modular Self-organizing Systems"
The beta version 2.0 was initially developed by Dr. M. Rybnik
The development was accomplished under supervision of Prof. K. Madani and Prof. A. Chebira
My enhanced version of Tree-like Divide to Simplify (T-DTS) ANN (AI/ML) lego-like tool used for any classification task which is grounded on the following ideas:
- The set of DB classification decomposers and the set of the end-tree-leaf classifiers (Rybnik's implementation in code + his T-DTS code skeleton)
- Increased library of classification task complexety estimators (Rybnik's and my contributions plus a lot of debug)
- Among them RBF Net like IBM ZISC(r)-036 RBF Net based estimator that allows us to build T-TDS on chip (my contribution)
- Extra loop that allows us to find sub-optimal classification complexity estimator from lib based on max entropy principle and define its optimal value for a concrete classification problem as there is no absolute value and no approach for end-user to guess apriory except try, check, fail & try again approach (my contribution)