README.md

Manifold-based Dimensionality Reduction Methods

This repository contains Python implementations of two manifold-based dimensionality reduction methods based on the following research papers:

1. Sparsity Preserving Projections with Applications to Face Recognition
2. Discriminant Sparse Neighborhood Preserving Embedding for Face Recognition

These implementations were reconstructed as part of my thesis work.

Research Paper

• Description: My thesis paper explores the implementation and applications of sparsity-preserving projections and discriminant sparse neighborhood preserving embedding for face recognition. It provides in-depth insights into the methods and their performance.
  • Link to My Thesis Paper (Greek)
  • Published in: IKEE AUTh (GRI-2023-41084)

Presentation

• Overview: This presentation offers a visual walkthrough of the key concepts, methodologies, and findings presented in my thesis. It serves as a supplementary resource to better understand the implemented dimensionality reduction methods.
  • Link to My Thesis Presentation (Greek)

Methods Implemented

1. Sparsity Preserving Projections (SPP):

   • Description: SPP is a dimensionality reduction method that preserves the sparsity structure of data, which is particularly useful for face recognition applications.
   • Implementation: SPP_method.py

2. Discriminant Sparse Neighborhood Preserving Embedding (DSNPE):

   • Description: DSNPE is a technique designed for face recognition that combines discriminant analysis with sparsity preserving projections.
   • Implementation: DSNPE_method.py

Usage

To use these dimensionality reduction methods in your Python projects, you can simply import the respective modules:

from SPP_method import SPP
from DSNPE_method import DSNPE

#SPP use
spp = SPP(n_components=50, epsilon=0.05)
spp.fit(X)
X_spp = dsnpe.transform(X)

#DSNPE use
dsnpe = DSNPE(n_components=50, epsilon=0.05, gamma=1.0)
dsnpe.fit(X, y)
X_dsnpe = dsnpe.transform(X)