Face Recognition Project

Overview

This project consists of two distinct parts.
The first part focuses on implementing Gradient Descent to optimize parameters for regression-like models and analyze cost function convergence.
The second part is dedicated to face recognition and reconstruction using Eigenfaces and Singular Value Decomposition (SVD), exploring dimensionality reduction and reconstruction quality.
Both parts include visuals, experiments and performance evaluation but they address separate objectives.

Highlights

Part 1 - Gradient Descent:

• Implementation of gradient descent for parameter optimization
• Use of fixerd and variable learning rates
• Analysis of cost function convergence over iterations
• Visualization of model predictions against training data
• Exploration of underfitting, overfitting and learning rate effects

Part 2 - Face Recognition and Reconstruction:

• Computation of Eigenfaces and singular vectors using PCA and SVD
• Reconstruction of faces and evaluation of reconstruction error
• Analysis of the effect of the number of principal components (d) on reconstruction quality
• Visualization of top eigenfaces, singular vectors, original images and reconstructions