There are two major advantages to quantum computing as such: the vast increase in computing power that comes from the quantum evolution of states, and the ability (at least in principle) of performing computations that cannot be done, classically. However major problems exist to their practical implementation. The marriage of quantum computing to artificial neural networks potentially solves three pressing problems in quantum computing: 1) the paucity of quantum computing algorithms; 2) the need to have continuous error correction to deal with decoherence and incomplete and/or damaged data; and 3) the practical difficulties of scaling up to large numbers of qubits. This marriage also addresses a major hurdle in neural networks: 4) that of supplying the connectivity for large scale neural networks.
Here the PI's propose an interdisciplinary collaborative research effort between experimental work with SQUIDs at the University of Kansas and theoretical work in quantum neural networks at Wichita State University. This research is directed toward building a quantum neural network to exploit the advantages both of quantum computers and of neural networks. There are two avenues they wish to explore. First, they will extend their work on temporal quantum neural networks, in which information flows in the temporal dimension (temporal network), to the case of using superconducting quantum interference devices (SQUIDs) as qubits. An experimental physical implementation of their temporal quantum neural network will be trained to learn complex behavior such as logic gates. Second, they will continue and expand their current work in theoretical characterization and simulation of a quantum neural networks in which information flows in the spatial dimension (spatial network), and extend their experimental work to the case of several entangled qubits.
