By NTT Staff
Our PHI Lab is focused on solving the most difficult computational problems ever conceived. To do so, they are employing a novel computing system called a Coherent Ising Machine (CIM). Three years ago, PHI Lab Director Yoshihisa Yamamoto and seven colleagues introduced the basic concept of the CIM and its quantum feature in Quantum Information, a Nature Partner Journal.
Experimental results have been promising, and the PHI Lab hopes to be able to share more recent results soon. One challenge they have encountered, however, is that under certain conditions – namely, when the laser pump is increased from below to above threshold – CIMs may be prevented from relaxing to their true ground state. In another article co-authored by Dr. Yamamoto and colleagues at Stanford University and the University of Tokyo, and published in the latest issue of Applied Physics Letter (APL), we learn of two possible solutions to that problem : 1) coherent spreading over local minima via quantum noise correlation; and 2) real-time error correction feedback.
In the course of discussing these approaches, the authors turn to a wide range of interdisciplinary concepts, which they describe as “spanning the fields of statistical physics, mathematics, and computer science, including dynamical systems theory, bifurcation theory, chaos, spin glasses, belief propagation and survey propagation.”
What is happening in this quest for explanations is not uncommon in pathbreaking scientific research. In this case, the high-level performances of CIM prototypes have simply outpaced our current levels of understanding. Interestingly, that stands in contrast to mainstream quantum computing, which to date has been characterized by a higher level of theoretical analysis than laboratory results. A reasonable objective for CIM research now is to advance both sides of the equation, even if the theoretical work has the effect of creating a field of study not listed in existing academic taxonomies.
“We look forward to accelerated advancement of learning in both the theoretical and experimental studies of CIMs … [and] see many rich possibilities for future interdisciplinary research, focused around a multifaceted theoretical and experimental approach to combinatorial optimization that unites perspectives from statistics, computer science, statistical physics and quantum optics,” Dr. Yamamoto said.
The PHI Lab has already cast a wide net in its long-range goal to radically redesign classical computers, through its numerous joint research agreements and its embrace of adjacent fields, such as biological neural networking. If we are at the dawn of a new field of study, it could be a while before we see academics codifying it or publishers launching new journals and conferences, but the APL article by Dr. Yamamoto and his colleagues appears to be a start.