CIFAR-10 Image Classification

This project explores deep learning architectures for image classification on the CIFAR-10 dataset using PyTorch. The goal is to classify images into one of 10 classes: airplane, automobile, bird, cat, deer, dog, frog, horse, ship, and truck.

---

Features

This repository contains a single Python script that handles:

1. Data Loading – Automatically downloads the CIFAR-10 dataset and prepares training, validation, and test splits.
2. Model Definitions – Includes linear, non-linear (fully connected with ReLU), and convolutional neural network (CNN) models.
3. Training and Evaluation – Trains models using SGD, evaluates on validation and test sets, and calculates accuracy.
4. Hyperparameter Search – Implements random and grid search to find top-performing model configurations.
5. Plotting – Generates plots of training and validation accuracy for the top hyperparameter configurations.

---

Skills

• PyTorch deep learning framework
• Convolutional Neural Networks (CNNs)
• Fully connected neural networks with ReLU activation
• Data preprocessing and batching with DataLoader
• Hyperparameter tuning (random search, grid search)
• Model evaluation, plotting, and visualization

---

Results for Top 3 Configurations for Different Network Architectures

Validation and training accuracy over number of filters, filter size and pooling size

Fully-connected output, 1 fully-connected hidden layer

Convolutional layer with max-pool and fully-connected output

---

Customization

Modify the train and parameter_search functions to experiment with different architectures, optimizers, or hyperparameter ranges.

Adjust SAMPLE_DATA for quick testing or full training.

Requirements

• Python 3.8+
• PyTorch 2.x
• torchvision
• matplotlib
• numpy
• tqdm

Installation

Install the dependencies with:

pip install torch torchvision matplotlib numpy tqdm

## Project Structure
cifar10_classification/
│
├─ cifar10_classification.py      # Main script (data, models, training, hyperparameter search, plotting)
├─ README.md                      # Project overview-- you're here :)
├─- `figures/`                    # plots of MSE results
└─ data/                          # CIFAR-10 dataset will be downloaded here automatically