notesum.ai

Published at November 25

Quantum Circuit Training with Growth-Based Architectures

Released Date: November 25, 2024

Authors: Callum Duffy¹, Smit Chaudhary², Gergana V. Velikova²

Aff.: ¹Centre for Data Intensive Science and Industry (DISI), University College London, Gower Street, London WC1E 6BT, United Kingdom; ²PASQAL, 7 Rue L´eonard de Vinci, Massy, France, 93100

Arxiv: http://arxiv.org/abs/2411.16560v1

	Mean Squared Error
Model	Mean	Best	Worst
5-layer CDL (RAND)	$(1.0\pm 4.9)\times 10^{-2}$	$1.76\times 10^{-6}$	$2.65\times 10^{-1}$
20-layer CDL (RAND)	$(8.13\pm 1.42)\times 10^{-5}$	$7.63\times 10^{-6}$	$1.0\times 10^{-2}$
5-layers CDL ( $\mathbb{I}$ )	$(4.8\pm 1.3)\times 10^{-2}$	$1.86\times 10^{-9}$	$4.35\times 10^{-1}$
20-layer CDL ( $\mathbb{I}$ )	$(5.5\pm 1.1)\times 10^{-5}$	$1.084\times 10^{-6}$	$3.37\times 10^{-4}$
Block growth	$(1.03\pm 3.52)\times 10^{-5}$	$1.46\times 10^{-7}$	$2.16\times 10^{-4}$
Seq FM Growth	$(1.57\pm 5.29)\times 10^{-5}$	$1.060\times 10^{-9}$	$2.87\times 10^{-4}$
Int FM Growth	$\boldsymbol{(1.029\pm 3.52)\times 10^{-6}}$	$\boldsymbol{8.889\times 10^{-10}}$	$\boldsymbol{1.259\times 10^{-5}}$