A multi-disciplinary Artificial Intelligence suite featuring core algorithms implemented in Java and MATLAB. This portfolio covers Search Strategies, Constraint Satisfaction, and Machine Learning.
The repository is divided into three specialized modules:
An interactive game implementing search strategies to manage game states and intelligent decision-making.
- Core Logic: State-space search and move validation.
- Features: Human vs. CPU interaction logic.
- Language: Java (Object-Oriented Programming).
A classic AI problem solved using backtracking to color a map such that no adjacent regions share the same color.
- Algorithm: Backtracking search with constraint checking.
- Logic: Efficiently assigning values to variables while respecting spatial boundaries.
- Language: Java.
A precision-focused module for pattern recognition and unsupervised grouping using a custom-curated dataset.
- Custom Digit Recognition: One-vs-All Logistic Regression trained and validated directly on 50 handwritten images (
0.pngto49.png). - Dynamic Data Loader: Features a custom script that transforms raw PNG pixel data into training matrices, achieving high accuracy for specific handwriting styles.
- K-Means Clustering: An unsupervised learning model that organizes the 50 images into clusters based on visual similarity, with real-time visualization of centroid convergence.
- Optimization: Utilizes the Conjugate Gradient (
fmincg) algorithm for efficient weight tuning. - Language: MATLAB.