Introduction and Overview

Quantum computing is a rapidly growing field, poised to address the enormous computational power demands of the future. The integration of machine learning (deep learning) into quantum machine learning (quantum deep learning) also holds the potential for significant value. In the current Noisy Intermediate-Scale Quantum (NISQ) era, due to the prohibitive cost and limited accessibility of real quantum computing resources, research in QML largely depends on classical computers to simulate quantum circuits. Therefore, with the boundless potential of this field, both domestic and international companies have already laid the groundwork in programming frameworks and software services, capitalizing on early-mover advantages to attract and cultivate potential users, even before the advent of fault-tolerant universal quantum computers.

In the era of deep learning, important considerations for a quantum computing framework include its simulation scale, the ability to leverage GPU power for efficiency, the degree of integration with machine learning/deep learning libraries, and its user-friendliness and convenience. Influential quantum computing frameworks often aim to provide real quantum computing resources to users through cloud platforms.

DeepQuantum is a lightweight quantum programming framework based on PyTorch, designed for programming and simulating quantum computing, quantum neural networks, and hybrid quantum-classical algorithms. It naturally integrates with PyTorch, offering a programming style closely aligned with PyTorch’s own. This makes it more accessible to developers with a computer science background and those familiar with or exposed to PyTorch, lowering the learning curve and easing the transition from machine learning (deep learning) to quantum machine learning (quantum deep learning). DeepQuantum is meticulously designed for convenient initialization of quantum neural networks and flexible data encoding. It also implements tensor network algorithms, supporting large-scale quantum circuit simulations based on matrix product states.