[Paper Review] TensorFlow Quantum: A Software Framework for Quantum Machine Learning

Michael Broughton, Guillaume Verdon+27 more|arXiv (Cornell University)|Mar 6, 2020

Quantum Computing Algorithms and Architecture13 references311 citations

TL;DR

TFQ is an open-source library that integrates Cirq with TensorFlow to enable rapid prototyping and training of hybrid quantum-classical models on classical or quantum data, with differentiable backpropagation through quantum circuits.

ABSTRACT

We introduce TensorFlow Quantum (TFQ), an open source library for the rapid prototyping of hybrid quantum-classical models for classical or quantum data. This framework offers high-level abstractions for the design and training of both discriminative and generative quantum models under TensorFlow and supports high-performance quantum circuit simulators. We provide an overview of the software architecture and building blocks through several examples and review the theory of hybrid quantum-classical neural networks. We illustrate TFQ functionalities via several basic applications including supervised learning for quantum classification, quantum control, simulating noisy quantum circuits, and quantum approximate optimization. Moreover, we demonstrate how one can apply TFQ to tackle advanced quantum learning tasks including meta-learning, layerwise learning, Hamiltonian learning, sampling thermal states, variational quantum eigensolvers, classification of quantum phase transitions, generative adversarial networks, and reinforcement learning. We hope this framework provides the necessary tools for the quantum computing and machine learning research communities to explore models of both natural and artificial quantum systems, and ultimately discover new quantum algorithms which could potentially yield a quantum advantage.

Motivation & Objective

• Motivate and enable rapid prototyping of hybrid quantum-classical models for quantum and classical data.
• Provide high-level abstractions for designing and training quantum models within TensorFlow/Keras.
• Bridge Cirq and TensorFlow to support differentiable quantum circuits and backpropagation.
• Offer high-performance quantum circuit simulation and practical demonstrations across applications.
• Foster exploration of advanced quantum learning tasks and potential quantum advantages.

Proposed method

• Introduce four design principles: differentiability, circuit batching, execution backend agnosticism, and minimalism.
• Represent quantum circuits and operators as tensors that feed into TensorFlow ops for simulation or hardware execution.
• Provide differentiable expectation-value computations and multiple gradient methods (finite difference, parameter shift, stochastic methods) for quantum circuits.
• Utilize a native high-performance simulator (qsim) and integrate Cirq objects via tfq.convert_to_tensor for seamless graph construction.
• Demonstrate a minimal hybrid model and scalable workflow from circuit construction to backpropagation within the TFQ stack.

Experimental results

Research questions

• RQ1 How can one design a software stack that enables differentiable, batched execution of parameterized quantum circuits within a classical ML framework?
• RQ2 Can Cirq-based quantum circuit concepts be effectively bridged with TensorFlow to support hybrid quantum-classical models in training and evaluation?
• RQ3 What are practical applications and workflows for applying TFQ to quantum data and hybrid learning tasks?
• RQ4 What strategies (e.g., gradient estimators, layerwise training) help mitigate training challenges in hybrid quantum-classical models?

Key findings

• TFQ enables differentiable backpropagation through hybrid quantum-classical models inside TensorFlow/Keras.
• It supports batched circuit execution, automated expectation estimation, and various gradient calculations for quantum circuits.
• TFQ provides a high-performance quantum circuit simulator (qsim) and hardware-agnostic execution backend options.
• The framework enables a range of applications, including quantum data classification, quantum control, and QAOA-like problems, as well as advanced tasks like meta-learning and reinforcement learning.
• TFQ architecture emphasizes leveraging existing Cirq and TensorFlow capabilities to minimize new learning curves for users.