AWS Quantum Technologies Blog

Working with PennyLane for variational quantum algorithms and quantum machine learning

The field of quantum computing today resembles the state of machine learning a few decades ago – in many ways. Near-term quantum algorithms for optimization, computational chemistry, and other applications are based on the very same principles that are used to train a neural network. In machine learning, there was no theoretical proof that a neural network could successfully recommend a good book to read. Similarly, little is known about the computational power white board sketch of variational algorithmsof variational quantum algorithms that iteratively adjust circuit parameters to arrive at a solution — much like a neural network. The field of quantum computing can learn a lot from the insights gained in the machine learning space over the past decades.

To bring the two fields of machine learning and quantum computing closer together, people in the quantum community began exploring how to extend one of the foundational concepts of machine learning — auto-differentiation — to quantum algorithms. These so-called parameter shift rules are at the core of what today is called quantum differentiable programming, and were first introduced by Mitarai et al. (2018) and shortly after extended in Schuld et al. (2018). The authors of the latter paper went on to create PennyLane, an open source project for variational quantum computing and quantum differentiable programming that allows you to build and ‘train’ quantum algorithms, like you would for a neural network. PennyLane lets you to directly use existing machine learning (ML) libraries, such as PyTorch, JAX, or TensorFlow, to build quantum algorithms and run them on different quantum computers or circuit simulators. This enables quantum computing researchers to benefit from the mature and comprehensive tooling that was developed over many years in ML, while also enabling machine learning experts to experiment with quantum computing using familiar tools and terminology.

Last year, AWS began working closely with the PennyLane team, and in December, Amazon Braket launched support for PennyLane. In February, AWS announced that it joined the PennyLane steering council to support the goal of building better tools for developers and researchers by combining ideas from ML and quantum computing. Working with AWS partner Xanadu, we want to grow PennyLane as an open, community-driven project. I hope we get many contributors from quantum, ML, and other fields to join the project and build with us.

Learn more about PennyLane and differentiable quantum programming on Amazon Braket in our PennyLane post on the AWS Open Source Blog.