Logistic Regression on MNIST (Digits 5 vs 6)

This project applies logistic regression to classify digits 5 and 6 from the MNIST dataset. It demonstrates:

• Data preprocessing
• Model training (with and without regularization)
• Evaluation with accuracy metrics and confusion matrices
• ROC curve plotting
• Hyperparameter tuning using GridSearchCV

🧠 Model Description

Two logistic regression models are trained:

• With L2 regularization
• Without regularization

The models are evaluated based on overall accuracy, per-class accuracy, and AUROC scores.

📊 Visualization

• Sample digits display (5 and 6)
• Accuracy vs. training sample size
• Accuracy vs. regularization strength (C)
• ROC curves

🧪 Requirements

Install the required libraries using:

pip install numpy matplotlib scikit-learn