Hybrid Neural Network with Qiskit and CNN

This project is to research quantum usage in AI field. We created hybrid Neural Network with quantum circuit (IBM Qiskit) and CNN, and tried the model in different tasks such as classifying grayscale number image in MNIST Dataset and colorful pets image in The Oxford-IIIT Pet Dataset.

Structure

Below is the main structure of the hybrid neural network.

Quantum Circuit with CNN

Datasets

We used handwriting number in MNIST Dataset and pets image in The Oxford-IIIT Pet Dataset.

MNIST Dataset

The Oxford-IIIT Pet Dataset

Tools

• PyTorch
• Qiskit
• Colab

Result

We used different amount of quantum in quantum circuit and compared the result in classifying cats and dogs of hybrid neural network with tanh and sigmoid function, found out the accuracy of hybrid neural network had no significant differences.

Handwriting Number Test Result

Website

• IBM Qiskit with Pytorch
• MNIST Dataset
• The Oxford-IIIT Pet Dataset
• Google Colab