Quantum learning machines

TL;DR

This paper explores the thermodynamic and energetic aspects of quantum learning machines, highlighting their efficiency and the fundamental differences in what is learned compared to classical machines.

Contribution

It introduces a quantum optical perceptron model demonstrating minimal energy dissipation and discusses how quantum measurement defines what is learned, unlike classical systems.

Findings

Quantum perceptron dissipates energy via spontaneous emission.

At optical frequencies, the perceptron operates near thermodynamic efficiency.

Quantum measurement determines the learned information, differing from classical objective facts.

Abstract

Physical learning machines, be they classical or quantum, are necessarily dissipative systems. The rate of energy dissipation decreases as the learning error rate decreases linking thermodynamic efficiency and learning efficiency. In the classical case the energy is dissipated as heat. We give an example based on a quantum optical perceptron where the energy is dissipated as spontaneous emission. At optical frequencies the temperature is effectively zero so this perceptron is as efficient as it is possible to get. The example illustrates a general point: In a classical learning machine, measurement is taken to reveal objective facts about the world. In quantum learning machines what is learned is defined by the nature of the measurement itself.