QCNN

Learnability of Noise and Error Mitigation in Hybrid Quantum-Classical CNN

This repository is the implementation of a Quantum CNN.

Installation

The following section describes the necessary steps to develop and execute your code with VSCode (an IDE specifically designed for Python) in a virtual machine within a Linux virtual environment.

How to use?

train.py trains a hybrid model on the specified settings in settings.yaml.

Inputs:

• Use one of the yaml files in ./settings as a baseline and change parameters as desired.
• Run python3 train.py --config-name <name of the yaml file in ./settings>, e.g. python3 train.py --config-name settings_2D_breastMNIST.yaml.'python3 train.py --config-name settings_2D_medMNIST.yaml'
• Multiple runs can be done in one command line using the MULTITUN argument, e.g. python3 train.py --config-name settings_2D_breastMNIST.yaml hydra.mode=MULTIRUN ++PARAMS.all_seeds=1,2,3,4,5 to run the same settings with five different seeds. See the hydra documentation if needed.

Outputs:

• SAVE_DIR/
  • SAVE_NAME/EXP_NAME/
    • circuit.txt #text file with a drawing of the quantum circuit
    • events.out.tfevents #tensorboard logs
    • train_val_curves.png #png image of the training and validation train_val_curves
    • progress.csv #csv summary of training and validation progress
    • checkpoints/ #checkpoints files

File Description

• calculations - custom_entanglements.py - Sequence_RX_CNOTs
• data - medmnist.py
• encoders - custom_higher_order_encoder.py
• settings - settings_2D_medMNIST.yaml

This Repo is still under development.