Fast optimal structures generator for parameterized quantum circuits

TL;DR

This paper introduces a rapid, meta-trained graph VAE-based method for optimizing quantum circuit structures in variational quantum algorithms, significantly reducing time and improving performance.

Contribution

It presents a novel meta-trained approach that automatically determines quantum gate numbers and generates optimal circuit structures for new tasks, outperforming existing algorithms.

Findings

Structures with lower loss generated

70 times faster than DQAS

Effective filtering of poor-performing circuits

Abstract

Current structure optimization algorithms optimize the structure of quantum circuit from scratch for each new task of variational quantum algorithms (VQAs) without using any prior experience, which is inefficient and time-consuming. Besides, the number of quantum gates is a hyper-parameter of these algorithms, which is difficult and time-consuming to determine. In this paper, we propose a rapid structure optimization algorithm for VQAs which automatically determines the number of quantum gates and directly generates the optimal structures for new tasks with the meta-trained graph variational autoencoder (VAE) on a number of training tasks. We also develop a meta-trained predictor to filter out circuits with poor performances to further accelerate the algorithm. Simulation results show that our method output structures with lower loss and it is 70 times faster in running time compared to a state-of-the-art algorithm, namely DQAS.