An approach of quantum learning machine to develop deterministic algorithms

Abstract

A novel approach of a quantum learning machine is introduced. A quantum learning machine can be used for automatically controlling quantum coherence and for developing new deterministic algorithms. Quantum learning machine contains the following devices: A preparation device P to prepare quantum system, an operation device U for unitary operation, a measurement device M performing quantum test and a feed-back system F. The feed-back system F is responsible for the learning process and it is equipped with classical memory of success/failure and a certain learning algorithm. In this work, we focus on finding quantum algorithms. The work is two-fold: First, we illustrate that quantum learning machine works with no a priori knowledge on its algorthm. As an example, we demonstrate that the quantum learning machine can learn Deutsch`s task so that it finds a quantum algorithm, different from but equivalent to the original one. Second, we show that it is always possible to complete the learning process in a finite time and to work more efficiently than a primitive feed-back model designed without any access to the classical memory, recording success/failure events.