Crushing runtimes in adiabatic quantum computation with Energy Landscape Manipulation (ELM): Application to Quantum Factoring

TL;DR

This paper presents two novel methods for accelerating adiabatic quantum computation by manipulating the energy landscape, demonstrated on quantum factoring, significantly reducing runtimes and improving efficiency.

Contribution

The authors introduce two general techniques for increasing energy gaps and transforming problem landscapes to speed up adiabatic quantum algorithms, applicable to various optimization problems.

Findings

Maximum runtime reduced by up to 754% in quantum factoring example

Methods successfully alter energy landscape without changing ground state

Techniques can be combined for further improvements

Abstract

We introduce two methods for speeding up adiabatic quantum computations by increasing the energy between the ground and first excited states. Our methods are even more general. They can be used to shift a Hamiltonian's density of states away from the ground state, so that fewer states occupy the low-lying energies near the minimum, hence allowing for faster adiabatic passages to find the ground state with less risk of getting caught in an undesired low-lying excited state during the passage. Even more generally, our methods can be used to transform a discrete optimization problem into a new one whose unique minimum still encodes the desired answer, but with the objective function's values forming a different landscape. Aspects of the landscape such as the objective function's range, or the values of certain coefficients, or how many different inputs lead to a given output value, can be decreased *or* increased. One of the many examples for which these methods are useful is in finding the ground state of a Hamiltonian using NMR: If it is difficult to find a molecule such that the distances between the spins match the interactions in the Hamiltonian, the interactions in the Hamiltonian can be changed without at all changing the ground state. We apply our methods to an AQC algorithm for integer factorization, and the first method reduces the maximum runtime in our example by up to 754%, and the second method reduces the maximum runtime of another example by up to 250%. These two methods may also be combined.