TensorFlow Quantum: A Software Framework for Quantum Machine Learning (2003.02989v2)

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.

• The paper introduces TensorFlow Quantum, a framework that merges classical neural networks with quantum circuits to build hybrid models.
• It demonstrates the use of parameterized quantum circuits and automatic differentiation to effectively train quantum-classical systems.
• The framework enables scalable simulation, supporting advanced applications such as quantum CNNs and noisy circuit analysis.

Overview of TensorFlow Quantum: A Framework for Quantum Machine Learning

The paper introduces TensorFlow Quantum (TFQ), a software framework designed to facilitate quantum machine learning research by bridging the gap between classical and quantum computational models. TFQ combines TensorFlow, a leading machine learning platform, with Cirq, Google's open-source library for developing quantum circuits, enabling seamless integration of classical machine learning models with quantum circuits.

Key Contributions

Integration and Differentiability:

• Quantum-Classical Models: TFQ allows the construction of hybrid models, where quantum circuits and classical layers coexist and communicate fluidly. This is facilitated by the parameterized quantum circuit (PQC) layer abstraction, which enables the formation of models combining classical neural networks with quantum computations.
• Automatic Differentiation: A core feature of TFQ is its support for differentiating quantum circuits. The capability to backpropagate gradients, a cornerstone of deep learning, is extended to quantum models, enabling effective training through gradient descent and similar methods.

Scalable Simulation:

• High-Performance Simulation: TFQ is equipped with qsim, a highly efficient simulator for executing quantum circuits. This simulator harnesses hardware acceleration and optimized algorithms to simulate circuits with a significant performance advantage over existing tools, such as Cirq’s native simulators.

Selected Applications

The paper illustrates several innovative applications of TFQ, showcasing its versatility:

• Quantum Convolutional Neural Networks (QCNNs): Leveraging the translational invariance in quantum data, TFQ enables the construction of QCNN architectures that paper phase transitions in quantum systems.
• Quantum Control: The framework is used to develop models for optimizing quantum control processes, crucial for calibrating quantum devices and mitigating errors in quantum computations.
• Noisy Circuit Simulation: TFQ aids in the simulation of quantum circuits under realistic noise models, which is essential for understanding the behavior and limits of quantum algorithms on noisy intermediate-scale quantum (NISQ) devices.

Advanced Research Directions

TFQ supports more sophisticated research through its high-level abstractions, including:

• Meta-Learning for Quantum Optimization: Innovative learning algorithms employ recurrent neural networks (RNNs) to optimize variational quantum circuits, demonstrating TFQ's capability to intertwine classical and quantum machine learning techniques.
• Hamiltonian Learning with Quantum Neural Networks: TFQ aids in learning Hamiltonian dynamics using quantum graph-based models, applying quantum neural networks to infer the structure and parameters of a quantum Hamiltonian from observed data.

Implications and Future Directions

TFQ's development marks a substantial step towards integrating quantum computing into mainstream machine learning workflows. By offering a platform where quantum circuits can be efficiently combined with classical deep learning architectures, TFQ empowers researchers to explore the full potential of quantum-enhanced algorithms. The implications of such advancements could be significant, offering new pathways in quantum chemistry, optimization, and beyond.

Future work may focus on expanding the framework's capabilities to support broader classes of quantum circuits and further integration with emerging quantum hardware. As the field of quantum machine learning evolves, TFQ is positioned to play a crucial role in bridging theoretical advancements with practical implementation, potentially accelerating the discovery of quantum-advantaged algorithms.