This repository hosts the accompanying software for the following master thesis.
Abstract: For gate-based quantum computing, the design of quantum circuits is a critical procedure for implementing specific quantum algorithms. Currently, many quantum circuits are designed using Quantum Architecture Search, where heuristic algorithms are often applied, resulting in circuits that are not human-interpretable. To understand the underlying logic and specific patterns of the quantum circuits, we have developed a QASM-to-Qiskit transformation called the quantum decompiler. This transformation acts as a form of reverse engineering, converting the relatively low-level representation of a quantum circuit – the QASM file into a more understandable, high-level representation, the Python Qiskit code. To implement this method, we combined Genetic Algorithms (GA) and Abstract Syntax Trees (AST). In our work, we primarily focus on developing the concept and testing this proof-of-concept on some simple, commonly used quantum circuits (GHZ, QFT, QPE) with a limited number of qubits. At the end of this research, the metrics used for the evaluation of the output of our decompiler is also discussed.
The Quantum Circuit Genetic Decompiler project provides a modular approach to evolving quantum circuits using genetic programming techniques. This project is split into 3 main components, each responsible for handling different aspects of the quantum circuit decompilation and optimization process.
- dataset.py: Manages the loading, preprocessing, and handling of quantum circuit datasets. This module ensures that data is formatted and ready for use in the decompilation process.
- q_ast_gen.py: Responsible for generating random quantum circuits. This module serves as the core for creating initial circuit conditions and potential solutions that the genetic programming can evolve.
- deqompiler.py: Contains the logic for the genetic algorithm, including mutation, crossover, and selection mechanisms tailored for optimizing quantum circuits. It is also the entry point of the application, orchestrating the flow between dataset management, circuit generation, and the decompilation process. It initializes the process, executes the genetic algorithm, and outputs the results.
If you find the repository useful, please consider citing:
@misc{deqompiler,
author={Xie, Shubing and Sarkar, Aritra},
title={DeQompiler: Genetic Programming based QASM-to-Qiskit decompiler},
howpublished={\url{[https://github.com/Advanced-Research-Centre/DeQompile](https://github.com/Advanced-Research-Centre/DeQompile)}},
year={2024}
}