Quantum Convolutional Classifier (QCC)

Installation

Create virtual environment [optional]

It is recommended to install this package in a virtual environment. One good option is mamba, an alternative to conda based on C++.

After installing, run the following in the main directory of the repo.

mamba create --prefix $PWD/.env python
mamba activate $PWD/.env

To build and install the qcc package, run the following.

pip install -e .[dev] # --editable installation

Execution

# List the various options for running QCC
qcc --help

# Run the test suite
cd test; qcc load config/

Code Structure

Source Code File Tree (src/qcc)

.
├── cli # Command-line interface (start here)
├── experiment # Generic class that runs multiple experimental trials
│   └── logger # Custom logging that stores outputs in DataFrames
├── file # Useful I/O commands
├── filters.py # Classical convolution and common filters
├── graph # Auto-formatted graphs based on plotly - a.k.a. David bot
├── ml # PyTorch implementations of ML models
│   ├── cnn.py # Convolutional neural network implementation
│   ├── data.py # Custom dataset wrapper
│   ├── ml.py # ML-related functions
│   ├── mlp.py # Multi-layer perceptron implementation
│   ├── model.py # Generic class for ML experiments (training + testing)
│   ├── optimize.py # Custom optimizer wrapper
│   ├── quantum.py # MQCC and MQCC Optimized
│   └── quanvolution.py # Quanvolution PyTorch module
└── quantum
    ├── pennylane # Quantum operations in Pennylane
    │   └── ansatz # Ansatz compatible with Pennylane
    ├── qiskit # Qiskit classes and functions 
    │   └── ansatz # Ansatz compatible with Qiskit
    │              # (WARNING: ML / Ansatz related items are likely jank)
    ├── quantum.py # Quantum-related functions
    └── qubits.py # 2D list for representing multidimensional quantum circuits

Docs / Examples File Tree (docs/)

TBD

Test Suite File Tree (test/)

TBD

Todo

• Write documentation / comments
  • Sphinx / MKDocs deployment?
• Use https://arxiv.org/pdf/quant-ph/0410184.pdf for shift operation
• Rewrite all draw methods to use plotly
• Rename variables to match publication(s) and report
  • Shift → stride
  • Permute → Data Rearrangement
  • Filter → Kernel
• Get Pennylane lightning working for faster execution
• Stride, dilation, padding in convolution
• Spin-off general methods into seperate KUARQ package
  • qcc.quantum
  • Qiskit gates
    • (Pennylane too?)
  • Experiment
  • Logger
  • Graph (plotly)
• Major project: research method of unifying Qiskit and Pennylane quantum gates
• There is a good chance I typed some Sequences as Iterables. Check this.
• Add bias terms using quantum
• Update config with better test set
• Why is mqcc_layer.ipynb returning different results than convolution.ipynb>
  • i think it is pennylane vs qiskit?