NTT, Japan’s National Institute of Informatics, and the University of Tokyo will be putting on-line what they call a “Quantum Neural Network” machine starting November 27th. This machine is not a general purpose gate-based quantum computer, but rather one designed for optimizing Ising models similar to the approach taken by D-Wave with their quantum annealing machine. There will not be any charge initially for access to this machine, although these organizations are intending on offering a commercialized version by 2020. The design of this machine is quite unique due to the following factors:
· The qubits are implemented using photonics with pulses of light that circulate in a 1Km fiber optic cable.
· This machine runs at room temperature and does not require a dilution refrigerator. As a result power consumption is only 1 Kilowatt.
· The machine will initially be implemented with 2000 qubits, but the design appears to be easily scalable to higher capacities with a goal of reaching 100,000 qubits by March 2019.
· The Ising optimization models require programming the level of interaction between different qubits and the connectivity between qubits is a very important factor. While the D-Wave machine currently is only able to connect each qubit to six or less neighboring qubits, there does not seem to be any connectivity limits in this “Quantum Neural Network” machine. This could potentially provide a large advantage for fitting in problems within the constraints of the machine.
As a result, the architecture of this photonic based “Quantum Neural Network” machine could theoretically have a large advantage over the superconducting quantum annealing approach used by D-Wave. However, we will need to see if this architecture can realize these advantages in actuality after researchers have had a chance to try it out. For more details on this machine you can read a technical paper (in Japanese) here and a video that explains the architecture here. After November 27th, you will be able to try out the machine for yourself by going to https://qnncloud.com.