Quantum computing with differentiable quantum transforms

TL;DR

This paper introduces a framework for differentiable quantum transforms that manipulate quantum programs while maintaining differentiability, enabling optimization and resource efficiency in quantum computing tasks.

Contribution

The paper presents a novel framework for differentiable quantum transforms implemented in PennyLane, allowing for parametrization, optimization, and improved resource management in quantum programming.

Findings

Transforms are differentiable and parametrizable.

Framework applied to gradient computation, circuit compilation, error mitigation.

Potential for resource optimization in quantum computing tasks.

Abstract

We present a framework for differentiable quantum transforms. Such transforms are metaprograms capable of manipulating quantum programs in a way that preserves their differentiability. We highlight their potential with a set of relevant examples across quantum computing (gradient computation, circuit compilation, and error mitigation), and implement them using the transform framework of PennyLane, a software library for differentiable quantum programming. In this framework, the transforms themselves are differentiable and can be parametrized and optimized, which opens up the possibility of improved quantum resource requirements across a spectrum of tasks.