Google Research

TensorFlow Quantum: A Software Framework for Quantum Machine Learning

(2020)

Abstract

We introduce TensorFlow Quantum, an open-source software platform for the rapid prototyping of hybrid quantum-classical parameterized models to learn from classical or quantum data. In particular, this framework facilitate training, inference, and testing of hybrid quantum-classical models by providing a number of features including batched circuit execution, automated expectation estimation, estimation of quantum gradients, hybrid quantum-classical autodifferentiation, and rapid model construction, all from within TensorFlow. We present an introduction to the software architecture and main building blocks through several examples and code snippets. We provide further theoretical background into the theory of hybrid quantum-classical neural-network-based models and hybrid autodifferentiation. Moreover, we provide a set of concrete quantum applications including supervised learning with quantum classifiers, adaptive layer-wise training strategies for quantum neural network training, machine learning for quantum control, meta-learning for quantum neural network training, quantum dynamics learning, and quantum-probabilistic generative modelling of mixed quantum states. We hope this library provides the necessary toolbox for quantum computing and machine learning research communities to control, verify, and better understand the state and dynamics of natural or artificial quantum systems, design novel models for quantum matter and chemical systems, and ultimately discover new quantum algorithms to observe potential quantum advantage on quantum processors.

Learn more about how we do research

We maintain a portfolio of research projects, providing individuals and teams the freedom to emphasize specific types of work