T-DTS (Tree-like Divide to Simplify) is a modular, AI/ML classification framework developed in Matlab 6. This repository presents version 3.0, which was enhanced and extended as part of my doctoral thesis:
"Contribution to the Study and Implementation of Intelligent Modular Self-organizing Systems" Link to Thesis
T-DTS is designed as a Lego-like neural network tool for solving classification problems using a tree-based decomposition method and problem complexity estimation.
- Hierarchical, tree-like breakdown of classification problems into simpler, manageable subspaces
- Modular integration of:
- Classification decomposers
- End-node classifiers
- Estimators of classification complexity
- Based on core principles developed by Dr. M. Rybnik (ResearchGate)
- Significantly extended and debugged in this version
1. RBF-Based Estimator (My Contribution)
- Inspired by IBM ZISC-036 (Zero Instruction Set Computing)
- Implements a Radial Basis Function (RBF) approach: RBF Net - Wikipedia
- Enables T-DTS logic to be adapted for on-chip deployment
2. Max-Entropy Search Loop (My Contribution)
- Automatically selects the most effective complexity estimator from the library
- Based on the principle of maximum entropy
- Solves the trial-and-error dilemma for selecting estimator parameters
- Replaces "try, check, fail, repeat" with guided selection logic
- Version 2.0 (Beta): Developed by Dr. M. Rybnik
- Version 3.0: Fully restructured and enhanced as part of my PhD research
Supervised by:
- Prof. K. Madani (ResearchGate)
- Prof. A. Chebira (ResearchGate)
- Classification in complex, high-dimensional datasets
- Experimental modular neural network design
- Dynamic complexity estimation
- Educational tool for ANN self-organization concepts
- Prototype logic for embedded, on-chip AI processing
t-dts, tree-like divide to simplify, modular neural networks, ai classification framework, self-organizing systems, radial basis function estimator, rbf net, ibm zisc, on-chip ai, complexity estimation, entropy-based model selection, matlab neural network, phd thesis ai, artificial intelligence research, object-oriented ai modeling, machine learning tree-based, modular ai architecture, neuro-inspired system design, classification algorithm tool, small footprint neural networks
If you're exploring modular AI systems, classification strategies, or experimental ANN architectures — this project can serve as a foundation or point of reference.
To support visibility of academic and modular AI work: Please consider giving this repository a star (?). It helps other researchers and developers discover underrepresented approaches to intelligent systems.