Mode connectivity in the QCBM loss landscape

TL;DR

This paper investigates the loss landscape connectivity of quantum circuit Born machines (QCBMs), comparing different parameterizations and entangling layer designs to understand how these choices affect model performance and landscape features.

Contribution

It introduces an analysis of loss landscape connectivity in QCBMs, highlighting how rotational gate choices influence the landscape and model optimization.

Findings

Rotational gate choices improve loss landscape connectivity.

Different ansatz designs show varied landscape features.

Model performance varies with parameterization and entangling layers.

Abstract

Quantum circuit Born machines (QCBMs) and training via variational quantum algorithms (VQAs) are key applications for near-term quantum hardware. QCBM ans\"atze designs are unique in that they do not require prior knowledge of a physical Hamiltonian. Many ans\"atze are built from fixed designs. In this work, we train and compare the performance of QCBM models built using two commonly employed parameterizations and two commonly employed entangling layer designs. In addition to comparing the overall performance of these models, we look at features and characteristics of the loss landscape -- connectivity of minima in particular -- to help understand the advantages and disadvantages of each design choice. We show that the rotational gate choices can improve loss landscape connectivity.