Why wait for quantum computers to be perfected "someday," when you can use Toshiba's quasi-quantum optimization algorithm on Microsoft's Azure cloud in 2021? It outperforms today's fledgling quantum computer speeds through the use of that proprietary algorithm on conventional digital computers accelerated with cloud GPUs [graphic processing units].
Alternatively, the quasi-quantum algorithms can run on Toshiba's FPGA [field programmable gate array].
"It will take a very long time for quantum computers to achieve the high performance of our [GPU and FPGA-powered] optimization solutions for large-size problems," said Hayato Goto, chief research Scientist at Toshiba Corp. in Japan. In fact, Goto said, "So far, no one has even been able to prove that any future quantum computer will be able to solve combinatorial optimization problems faster, which leaves room for our classical machines to surpass quantum computers."
Too good to be true? Toshiba recently demonstrated a discrete version of its quasi-quantum algorithm in peer-reviewed benchmarks against a wide variety of current-day quantum computer hardware and quasi-quantum software simulators, and outperformed them all.
Toshiba says its newest quasi-quantum algorithms are available in two varieties, ballistic and discrete, the latter of which yields the fastest optimization solutions in the world, compared to hardware quantum annealers using superconducting circuits (such as those used in D-Wave Systems quantum computers), optically actuated quantum coherent Ising machines (used in Microsoft, Hewlett Packard, and Caltech/NTT quantum hardware), specialized digital optimization accelerators (such as Fujitsu's Digital Annealing Unit), and restricted Boltzmann machines (such as Google's generative stochastic artificial neural network).
The key to its high-speed large-problem optimization solutions, according to Toshiba, is the use of simulated bifurcation (SB), which is digitally accelerated with GPUs or FPGAs, both of which outperform both traditional digital optimization algorithms on CPUs (central processing units) and current quantum bifurcation hardware that emulates thermodynamic adiabatic annealing (as in D-Wave's hardware).
"Toshiba's simulated bifurcation algorithm is promising for a solver of optimization problems, especially its high-performance implementations using GPUs and FPGAs, " said processor research team leader Kentaro Sano at Japan's independent government-funded Institute of Physical and Chemical Research (RIKEN). "I also think that it can be favorably compared with present hardware of quantum [D-Wave] or digital [Fujitsu] annealers."
What's Quantum Bifurcation?
Quantum bifurcation is a qualitative—as opposed to quantitative—measurement that can be fine-tuned by the parameters governing the topological structure of quantum computers. An example is using a cavity bifurcation amplifier to fine-tune the "sweet spot" that maximizes the length of coherence in a quantronium (a Cooper-pair qubit). Toshiba's quasi-quantum bifurcation algorithm is a software simulation that manifests tunable qualities similar to the behavior of classical computers, according to the Encyclopedia of Complexity and Systems Science. The main advantage of Toshiba's digital quasi-quantum bifurcation algorithm is that optimization problems can be executed in parallel to achieve faster speed, and larger problem size, in proportion to the number of processors used.
Said Sano, "Toshiba's newest algorithm is capable of achieving both high parallelism and the ability to solve large-scale problems. Thanks to this, it can deliver the world's highest level of performance for optimization problems by using commercial processors without any special quantum hardware. Toshiba's algorithm will also allow increasingly higher performance in solving optimization problems to be achieved on newer, faster digital processors immediately after they become available."
Its greatest limitation, according to Sano, is that the simulation algorithms are specialized for optimization problems—which is why Toshiba benchmarked against quantum and digital annealers (which also specialize in optimization problems).
On the other hand, this limitation can also be viewed as a feature, since optimization problems in portfolio management, drug discovery, and logistics typically involve too many variables to be solved by humans, according to Sano.
"Optimization aims to find solutions as fast as possible from an enormous number of options—more than humans can fully explore," said Sano. "Deep learning and other gate-level problems, on the other hand, aim at making it possible for machines to do what humans already do, including image recognition and speech recognition."
As benchmarks go, quasi-quantum bifurcation can solve absurdly large problems—such as the world's largest optimization problem ever solved—a million bits—in just 30 minutes. The world's largest emulated annealers, in contrast, can only handle 5,000-bit problems. Current benchmarks (of latest-generation 2,000-bit emulated annealers) show hardware quantum annealers to be at best a tenth the speed of Toshiba's quasi-quantum bifurcation algorithm running on a fast multi-core digital computer. Without Toshiba's quasi-quantum bifurcation algorithm, even the world's highest-performance digital computers (supercomputers) are 20,000 times slower, taking as long as 14 months of 24/7 computing to solve the million-bit optimization problem, according to Toshiba senior research director Kosuke Tatsumura, as cited in Science Advances.
"We also expect further improvements in our algorithms," said Goto, "and further development of digital parallel computing technologies, both of which will accelerate our quasi-quantum bifurcation solution."
R. Colin Johnson is a Kyoto Prize Fellow who has worked as a technology journalist for two decades.