Hybrid-Quantum-Classification

Overview

This repository explores hybrid quantum–classical classification models, combining the strengths of classical deep learning with quantum machine learning.
The goal is to implement and evaluate different classification tasks using hybrid architectures where quantum circuits are embedded as trainable layers inside classical neural networks.

Hybrid models are particularly useful in researching:

• Near-term quantum algorithms (NISQ era)
• Quantum feature extraction
• Variational Quantum Circuits (VQCs) integrated with classical optimization

This repository serves as both a research playground and a learning resource for hybrid quantum machine learning.

Features

• ✅ Hybrid quantum–classical models using PyTorch + PennyLane
• ✅ Multiple classification tasks (binary and multi-class)
• ✅ Variational Quantum Circuits (VQCs)
• ✅ End-to-end training with classical optimizers
• ✅ Modular and extensible codebase
• ✅ CPU-based quantum simulators (no quantum hardware required)

Structure

Hybrid-Quantum-Classification/
│
├── data/
│   ├── datasets.py          # Dataset loading and preprocessing
│
├── models/
│   ├── classical.py         # Pure classical baselines
│   ├── quantum.py           # Quantum circuit definitions
│   ├── hybrid.py            # Hybrid quantum-classical models
│
├── training/
│   ├── train.py             # Training loop
│   ├── evaluate.py          # Evaluation metrics
│
├── experiments/
│   ├── binary_classification.ipynb
│   ├── multiclass_classification.ipynb
│
├── utils/
│   ├── config.py            # Hyperparameters and configs
│   ├── metrics.py           # Accuracy, loss, etc.
│
├── requirements.txt
├── README.md
└── main.py

Hybrid Model Architecture

A typical hybrid model in this repository follows this pipeline:

Classical preprocessing layer (PyTorch)

Quantum layer

Data encoding into quantum states

Variational quantum circuit (trainable parameters)

Classical post-processing layer

Loss computation & optimization using classical optimizers

Input → Classical Layer → Quantum Circuit → Classical Layer → Output

Classification Tasks

Implemented tasks include:

🔹 Binary classification (e.g., toy datasets, synthetic data)

🔹 Multi-class classification

🔹 Quantum-enhanced feature learning

🔹 Comparison with purely classical baselines

Each task is designed to highlight how quantum layers influence performance and learning behavior.

Installation

Install Dependencies

pip install -r requirements.txt

Results

Training curves and accuracy metrics are logged

Jupyter notebooks in experiments/ provide visual analysis

Classical vs hybrid performance comparisons included

Future Work

🔹 More expressive quantum ansätze

🔹 Larger datasets and feature maps

🔹 Quantum convolutional layers

🔹 Benchmarking against state-of-the-art classical models

License

This project is licensed under the MIT License. Feel free to use, modify, and distribute for research and educational purposes.