Machine Learning by Adiabatic Evolutionary Quantum System

TL;DR

This paper explores a quantum machine learning model called AEQS, which uses adiabatic evolution and quantum automata to solve relational problems efficiently through quantum algorithms.

Contribution

It introduces a method to train AEQSs for machine learning tasks using quantum algorithms and estimates their efficiency.

Findings

AEQSs can be controlled by 1qqaf's to solve relational problems.

Quantum algorithms like counting and amplitude estimation are utilized.

Preliminary efficiency estimates for quantum learning algorithms are provided.

Abstract

A computational model of adiabatic evolutionary quantum system (or AEQS, pronounced "eeh-ks") was introduced in [Yamakami,2022] as a sort of quantum annealing and its underlying input-driven Hamiltonians are generated quantum-algorithmically by various forms of quantum automata families (including 1qqaf's). We study an efficient way to accomplish certain machine learning tasks by training these AEQSs quantumly. When AEQSs are controlled by 1qqaf's, it suffices in essence to find an optimal 1qqaf that approximately solves a target relational problem. For this purpose, we develop a basic idea of approximately utilizing well-known quantum algorithms for quantum counting, quantum amplitude estimation, and quantum approximation. We then provide a rough estimation of the efficiency of our quantum learning algorithms for AEQSs.