While most of the attention on quantum computing currently tracks gate-based approaches, there has also been steady progress with Coherent Ising Machine (CIM) systems, which are essentially analog systems that seek to leverage quantum effects. It is the approach D-Wave uses. However, unlike the semiconductor-based superconducting technology that D-Wave employs, other technologies including optical approaches to CIM are possible and also rapidly developing.
Today, NTT Research and the Tokyo Institute of Technology announced an effort to collaborate on developing applications for CIM systems that highlights recent CIM progress and aspirations. NTT Research is a Sunnyvale, CA-based division of Japan’s telecom giant NTT’s R&D Lab. Its Physics and Informatics (PHI) Lab director, Yoshihisa Yamamoto, is a prominent CIM researcher with ties to Tokyo Tech and Stanford and NTT.
The focus of the new collaboration’s is on drug discovery and compressed sensing applications. As described in today’s official announcement, “[The] two agreements, signed in 2020, call for collaboration between NTT Research’s Physics & Informatics (PHI) Lab and independent research groups in Tokyo Tech’s School of Computing, directed by Yukata Akiyama and Toru Aonishi. NTT Research will lead the five-year project, which will involve approximately ten researchers working in Tokyo and Sunnyvale.” This agreement follows an NTT Research agreement announced in January with Caltech to develop and demonstrate “the world’s fastest CIM.”
It’s worth noting that NTT Research has ambitious goals:
“As part of its long-range goal to radically redesign artificial computers, both classical and quantum, the NTT Research PHI Lab has already established joint research agreements with seven universities, one government agency and one quantum computing software company. The other institutions of higher education are Cornell University, Massachusetts Institute of Technology (MIT), Stanford University, California Institute of Technology, Swinburne University of Technology, the University of Michigan and the University of Notre Dame. The government entity is NASA Ames Research Center, and the private company is 1QBit. In January 2021, NTT Research entered a second agreement with Caltech to develop an extremely fast, miniaturized CIM. The PHI Lab’s research partners include more than a dozen of the world’s leading quantum physicists. In addition to its PHI Lab, NTT Research has two other divisions: its Cryptography & Information Security (CIS) Lab and Medical & Health Informatics (MEI) Lab.”
The CIM approach to quantum computing is hardly new. In recent years, work to implement CIMs with optical technology has ramped up significantly. Here’s a broader CIM overview taken from a 2019 Nature[i] paper.
“To speed up calculation time compared to digital hardware, different non-von Neumann architectures have been proposed that attempt to solve optimization problems by mapping them to Ising models. Finding the optimal solution then becomes equivalent to finding the ground state of the Ising model, which is implemented with networks of coupled artificial Ising spins that can be realized with various physical systems, e.g. Josephson junctions, trapped ions, or optical states. The energy function of these so-called Ising machines is proportional to the Ising Hamiltonian, so that they will naturally evolve to the ground state of the Ising model and thus to the optimal solution. As the evolution to the ground state typically occurs on very fast timescales, Ising machines promise a considerable speed up over conventional algorithms in finding solutions to optimization problems, which will have significant implications for various important areas such as finance, pharmaceutics, logistics, or machine learning.”
As noted by Yamamoto in today’s announcement, much of the prior optically-based CIM work focused on understanding how quantum oscillator networks solve general combinatorial optimization problems. “Through this new application-oriented work undertaken in collaboration with Professors Akiyama and Aonishi, we believe that we will be able to explore new ways to use the networks by better understanding the requirements of a CIM,” said Yamamoto in the announcement. (For a visual representation of how a CIM solves a combinatorial optimization problem, see this video from the MIT’s Lincoln Laboratory.)
“The near-term goals in this joint research include formulating the essential part of the intensive computation required for a CIM to screen drug candidate compounds via combining their functional fragments and developing a CIM-based L0 norm reconstruction algorithm of distorted images. (The L0 norm relates to non-zero elements in a matrix.) Broader expectations are to demonstrate the advantages of a CIM and its related technology in addressing real-world problems and to explore new ways of computing,” according to NTT Research.
For more information see Yamamoto and colleagues’ paper (Coherent Ising machines—Quantum optics and neural network Perspectives) published in AIP last fall.
[i] A poor man’s coherent Ising machine based on opto-electronic feedback systems for solving optimization problems, 2019, https://www.nature.com/articles/s41467-019-11484-3