Heart Disease Identifier (from scratch)

Implemented logistic regression with GD/SGD/MBGD, plus a clean preprocessing pipeline.

• Vectorized updates, label encoding, feature scaling.
• Tracked loss per epoch; plotted ROC curve.
• Numerical stability fixes and reproducible runs.
• Preformed KNN and CART methodology.

Python · NumPy · matplotlib

Optimizer Study: GD vs SGD vs Mini-Batch

Compared convergence speed and generalization across optimizers on a logistic-regression task.

• Trained logistic regression (with and without L2 regularization).
• Ran GD, SGD, and Mini-Batch GD; tuned learning rate and batch size.
• Logged loss per epoch and plotted learning curves; reported train/test accuracy (and ROC AUC if applicable).

Python · NumPy · matplotlib

Computer Vision: Template Matching & Circles

Multi-scale template matching with SAD/NCC and robust Hough Circle detection.

• Detected small/medium/large templates across scales.
• Benchmarked SAD vs NCC; reported best match per noise level.
• Tuned Hough params to improve overlapping circle recall.

Python · OpenCV · matplotlib