Handwritten-digits-Classification

Overview

Hand written digit recognition using Logistic Regression, kernel SVM with PCA/LDA dimensionality reduction, and Deep Neural Network (Lenet-5 architecture) for MINST dataset.

To run the code

Deep Learning

python3 main.py --Method Lenet

• Parameters
  • Method - Classifiers. Default :- 'Lenet'

Logistic Regression

python3 main.py --Method LR --DimRed LDA 

• Parameters
  • Method - Classifiers. Default :- 'Lenet'
  • DimRed - Dimensionality Reduction technique. Option : 'PCA/LDA' Default :- 'PCA'

SVM

python3 main.py --Method SVM --DimRed LDA --Kernel Polynomial

• Parameters
  • Method - Classifiers. Default :- 'Lenet'
  • DimRed - Dimensionality Reduction technique. Option : 'PCA/LDA' Default :- 'PCA'
  • DimRed - kernel for Kernel SVM. Option : 'Polynomial/RBF' Default :- 'Linear'

Results

LeNet-5

Accuracy	Confusion Matrix

Logistic Regression

Dim Red.	Accuracy	Confusion
PCA
LDA

SVM

Kernel	PCA	LDA
Linear
Polynomial
RBF