Toshiba has invented an algorithm that it says delivers a 10-fold improvement for a select class of computational problems, without the need for exotic hardware. In fact, the company’s simulated bifurcation algorithm is said to offer advantages akin to a quantum computing system, while running on standard digital servers.
Developed by Toshiba researchers Hayato Goto and Kosuke Tatsumura, the new technique “harnesses bifurcation phenomena, adiabatic processes, and ergodic processes in classical mechanics” to rapidly find highly accurate approximate solutions for complex large-scale combinatorial optimization problems.
“These are problems that have resisted solution for a long time, and that are very difficult to solve using conventional techniques,” said Toshiba. “Potentially even more important, the algorithm also realizes excellent scalability at a low cost using current computers, which could revolutionize current optimization processes.”
First announced by Toshiba earlier this year, Bloomberg brought the story to the front page this week. The claimed breakthrough is being met with equal parts interest and skepticism from the financial services communities, where the technology offers serious potential.
Toshiba says it has developed a system capable of executing arbitrage opportunities for currencies in 30 microseconds. Fast enough to make profitable trades 90 percent of the time, Toshiba claimed. The company, working with financial services professionals, expects to complete a real-world trial by early 2021, according to Bloomberg.
Toshiba’s interest extends beyond that of an IT provider. According to a November 2019, news report from Nikkei, the Tokyo-based conglomerate is seeking to launch its own high-frequency trading business, where it will prove out the technology before selling it to financial institutions.
Earlier this year, Toshiba offered examples of how the highly parallelizable algorithm could achieve speedups using current hardware through parallel computation. Field-programmable gate arrays (FPGAs) could be leveraged to provide a good solution to an optimization problem with 2,000 fully connected variables (approximately 2 million connections) in 0.5 milliseconds, offering a 10 times speedup over a laser-based quantum computer, according to Toshiba. Further, the company said an eight-GPU cluster would be able to solve a large-scale problem involving 100,000 fully connected variables (about 5 billion connections) in a few seconds.
The algorithm excels at sifting through a vast number of options to find the best combinations. Aside from applications in finance, Toshiba highlights potential use cases in the areas of logistics, traffic management and drug development using molecular design.
These are the type of optimization problems for which quantum computing is considered a holy grail, but while quantum computing is advancing at a quickening pace, there’s always the chance that an algorithm could be discovered that could negate some of quantum’s advantage. While tech pundits debate the state of so-called quantum supremacy, Toshiba is betting that its quantum-inspired classical platform will at least in the near-term win out.
Science advances paper — https://advances.sciencemag.org/content/5/4/eaav2372