Physical realizations of a temporal quantum neural computer

Abstract

A single two-level system (TLS) evolving in time can act as a quantum neural computer. The system is prepared in some initial state (input(s)), and measured in some final state (output(s)). The states of the system at intermediate times are the (virtual) neurons, connected to each other by the time evolution operator and the influence functionals of the environment. The environmental variables coupling the system to itself at different times are trained using a standard gradient descent neural network algorithm. We demonstrate this by simulating first a quantum dot molecule on a solid state substrate, in which we tickle the substrate phonons, and second a variable /spl beta/ rf SQuID coupled to a microwave cavity field. Each can be trained to perform any desired logic gate, starting from a single initial physical state, and is thus a universal quantum computer.