Important Announcement: QNNcloud will be finished
Thank you for being interested in our QNNcloud. As our ImPACT program will be finished in next March, our QNNcloud web service will be finished. The schedule is as followings:
2018.12.28 (17pm Japan time): Acceptance of new calculation request will be finished. The other functions, such as viewing past results, are still avialble.
2019.mid March: QNNcloud web site will be closed.
We thank all of you for your kind support and understanding.
2018.11.1We have updated QNNcloud website. This updating includes the following changes:
・Improvement of the web site structure so that users can easily find directions of use.
・Limitaion of the number of times for use has been eased to 100 times per month.
・Addition of the usage guides (GUIDE, parameter information, and FAQ).
- 2018.5.30We have completed the maintenace of QNN machine.
- 2018.4.23We are going to have a maintenace of QNN machine from 2018.4.26 Thu.10am to 2018.5.27 Mon, noon in Japan time. During this maitenace, a new request for the QNN machine will not be accepted, but the other functions, such as login, checking your results, and viewing the other QNNcloud web pages will be available. We are sorry for any incovinience that may cause, and appreciate your understanding and cooperation.
- 2017.11.27 QNNcloud is now opened.
- 2017.11.27 QNN simulator will be coming soon.
Collective computing at criticality of OPO phase transition
Quantum Neural Network
What is QNNcloud?
QNNcloud is a cloud service that enables you to use a Quantum Neural Network (QNN), a new type of computer. It enables you to experience QNN computing with actual QNN computer hardware without having to be an expert in adjusting experimental optical equipment. It also enables you to learn about QNN operation principles through a quantum simulator that models the quantum-mechanical behavior of a QNN.
The basic elements of a QNN are quantum neurons that can that can enter a linear superposition state and quantum-measurement feedback circuits that serve as synapses interconnecting those neurons. A QNN searches for problem solutions using the phenomenon of phase-transition criticality. QNNcloud enables users to experience this new type of computer via the Internet. It can provide users in academia with the opportunity to verify QNN operation principles and performance and users in industry with the opportunity to develop algorithms for solving real-world problems. At present, users can experience the solving of difficult large-scale Max-Cut combinatorial optimization problems with a maximum of 2000 elements and all-to-all connections on actual QNN hardware.
Targets of QNNcloud
QNNcloud aims to bring QNN to general users as a new type of computer that has the potential of obtaining solutions to a wide variety of optimization problems at dramatically higher speeds than existing computers. In this way, QNN Cloud aims to release the latent power of QNN computing.
Today, there is a need in society for developing computers that can solve large-scale combinatorial optimization problems at high speeds in a variety of fields including lead compound optimization in the development of medicine, frequency band and/or transmission power optimization in wireless communications, sparse coding for compressed sensing, Boltzmann sampling in machine learning, and portfolio optimization in Fin Tech. In combinatorial-optimization and continuous-optimization problems, a “combinatorial explosion” can occur in which the number of optimal-solution candidates become huge as the scale of the problem increases. This problem cannot be handled by even quantum computers that are capable of parallel computing using quantum superposition much less by modern-day computers. The need has therefore arisen for establishing a new computing concept that can achieve massively parallel computing.
The QNN computer hardware solves optimization problems using the quantum mechanical behavior of a new type of laser called an optical parametric oscillator (OPO). This hardware achieves mutual coupling corresponding to the problem to be solved between many OPO pulses circulating in an optical fiber-ring cavity through the use of quantum-measurement feedback circuits.
More than 2000 OPO pulses are generated simultaneously in a 1-km long fiber ring cavity using a pump pulse train with a 1-GHz repetition frequency and a periodically poled lithium niobate (PPLN) waveguide. Coupling between OPO pulses is achieved by using homodyne detection to read a part of each OPO pulse sequentially removed from the cavity, calculating the magnitude of mutual coupling by a field programmable gate array (FPGA), and feeding back the result to the cavity. After many round trips, OPO pulses initially prepared in a 0-phase and π-phase superposition below the oscillation threshold search for the 0-phase or π-phase superposition that is most stable overall. This process derives a solution to the problem. [Further Reading => Technical Papers]
The QNN Cloud is the first step toward an application development platform using QNN computer hardware and is simultaneously a free service making this a particularly important step. With this system, we seek to uncover new application fields and algorithms together with QNN Cloud users around the world who have an interest in development.
The QNNcloud is researched and developed and operated by the National Institute of Informatics (NII) in cooperation with NTT under financial support of the Impulsing Paradigm Change through Disruptive Technologies Program (ImPACT).
The research and development of QNNs is being performed by NTT, NII, The University of Tokyo, Osaka University, Tohoku University, and Tokyo University of Science centered about a NII and Stanford University research team led by program manager Yoshihisa Yamamoto and supported by the Funding Program for World-Leading Innovating R&D on Science and Technology (FIRST) and ImPACT national projects.
Note: Inquiries on news releases should be directed to “Media Contacts” included in news-release material.
Experience the cutting-edge
Sign up right now?Machines Guide
In MACHINES, you can experience the solving of Max-Cut problems up to 2000 elements using actual QNN hardware.