mnist

MNIST Number Classification with ANN (PyTorch)

This project implements a simple Artificial Neural Network (ANN) using PyTorch to classify handwritten digits from the MNIST dataset.

The goal is to build an end-to-end deep learning pipeline including:

• Data loading and preprocessing
• Model definition
• Training
• Evaluation

---

Dataset

MNIST

• 60,000 training images
• 10,000 test images
• Image size: 28×28 (grayscale)
• 10 classes (digits 0–9)

The dataset is automatically downloaded using torchvision.datasets.MNIST.

---

Model Architecture

The model is a fully connected feed-forward neural network:

---

Training Details

• Framework: PyTorch
• Loss Function: CrossEntropyLoss
• Optimizer: Adam
• Learning Rate: 0.001
• Batch Size: 64
• Epochs: 10

---

Results

After training for 10 epochs:

Metric	Value
Final Loss	~0.065
Test Accuracy	97.13%

Training loss decreases steadily, showing stable convergence.

---

Sample Output

The notebook also visualizes sample digits from the dataset:

• Displays random MNIST images
• Shows their true labels
• Plots training loss per epoch

---

How to Run

1. Install dependencies:

pip install torch torchvision matplotlib

What This Project Demonstrates

This project shows:

• Basic PyTorch workflow
• GPU/CPU device handling
• Training and evaluation loops
• Clean modular code structure
• It is a solid baseline for more advanced models such as CNNs.

Possible Improvements

Future extensions could include:

• Adding convolutional layers (CNN)
• Using dropout and batch normalization
• Hyperparameter tuning
• Confusion matrix visualization