Facta Univ. Ser.: Elec. Energ., vol. 20, No. 3, December 2007, pp. 561-586.

Inductive Learning of Quantum Behaviors

Martin Lukac and Marek Perkowski

Abstract: In this paper studied are new concepts of robotic behaviors - deterministic and quantum probabilistic. In contrast to classical circuits, the quantum circuit can realize both of these behaviors. When applied to a robot, a quantum circuit controller realizes what we call \emph{quantum robot behaviors}. We use automated methods to synthesize quantum behaviors (circuits) from the examples (examples are cares of the quantum truth table). The don't knows (minterms not given as examples) are then converted not only to deterministic cares as in the classical learning, but also to output values generated with various probabilities. The Occam Razor principle, fundamental to inductive learning, is satisfied in this approach by seeking circuits of reduced complexity. This is illustrated by the synthesis of single output quantum circuits, as we extended the logic synthesis approach to Inductive Machine Learning for the case of learning quantum circuits from behavioral examples.

Keywords: Quantum circuits, machine learning, logic synthesi, quantum robot behavior.

16lukac