mosaiQue

A Python library for applying quantum convolutions with PennyLane, designed to facilitate image processing and transformation through custom convolution layers and quantum computations.

Table of Contents

• Installation
• Usage
• API Reference
• Contributing
• License

Installation

To install the library, clone the repository and install the required dependencies including conda-build

use develop to install the library

sudo /opt/conda/bin/conda-develop -n QML-QPF PATH /workspaces/QML-QPF/mosaiQue

Usage

Overview of the main classes in the library:

ConvolutionLayer4x4

• Create a convolution layer with a custom kernel shape.
• Fit the kernel to your dataset.
• Transform the dataset using convolution.
• Save a portion of the dataset or load a previously saved dataset.

QuantumLayer

• Define a quantum kernel operation using PennyLane.
• Create an operation layer using the quantum kernel.
• Preprocess images using a Keras model that incorporates the quantum layer.

Kernel2d4x4

• Initialize the kernel with a specific shape.
• Fit the kernel to the image data.
• Transform the image data into patches and reconstruct the original image data from patches.
• Merge channels back to the original shape.

API Reference

ConvolutionLayer4x4

Attributes

• name (str): The name of the convolution layer.
• kernel (kernels.Kernel2d4x4): The kernel used for convolution operations.

Methods

• fit(dataset: np.ndarray): Fits the kernel to the provided dataset.
• transform(dataset: np.ndarray) -> np.ndarray: Applies the convolution transformation to the dataset.
• post_transform(dataset: np.ndarray) -> np.ndarray: Applies post-processing to the dataset after transformation.
• channel_merge(dataset: np.ndarray) -> np.ndarray: Merges channels in the dataset.
• save(dataset: np.ndarray, variant: List[int]): Saves a portion of the dataset to a .npy file based on the variant.
• open(variant: List[int]) -> np.ndarray: Loads a saved dataset from a .npy file based on the variant.

QuantumLayer

Attributes

• q_node (qml.QNode): The quantum node to execute quantum computations.

Methods

• call(inputs: tf.Tensor) -> tf.Tensor: Executes the quantum layer on the input tensor.

OperationLayer

Attributes

• q_layer (QuantumLayer): The quantum layer for processing inputs.
• kernel_operation (qml.QNode): The QNode for the kernel operation.

Methods

• pre_op() -> keras.Model: Initializes and returns a custom Keras model used to preprocess images.

Kernel2d4x4

Attributes

• input_shape (list[int]): The shape of the input data.
• kernel_shape (list[int]): The shape of the kernel.

Methods

• fit(X: np.ndarray): Fits the kernel to the input data by storing its shape.
• transform(X: np.ndarray) -> np.ndarray: Splits the input data into smaller patches based on the kernel shape.
• post_transform(X: np.ndarray) -> np.ndarray: Reconstructs the image data from patches.
• channel_merge(X: np.ndarray) -> np.ndarray: Merges channels to restore the original order.

Contributing

Contributions are welcome! Please feel free to submit issues or pull requests. To contribute:

1. Fork the repository.
2. Create a new branch for your feature or bug fix.
3. Make your changes and commit them.
4. Push to your branch and create a pull request.

License

This project is licensed under the GPL-3.0 License. See the license file for details.