‘Optical Oracle’ Could Quickly Solve Complex Computing
SINGAPORE, SOUTHAMPTON, England, and MADRID, April 1, 2014 — The global optical fiber network may contain more computing potential than previously realized. Tapping into this could make such networks hundreds of times faster.
A team from Nanyang Technological University, the University of Southampton and IQFR-CSIC made the discovery. Acting as an “optical oracle,” the global network has demonstrated the ability to solve the Hamiltonian path problem, which is determining whether a route exists between multiple locations, or nodes, so that each is visited only once.
(Left) The optical oracle’s approach to solve the Hamiltonian path problem on a network with five nodes. (Right) An actual design of the optical oracle with optical fiber components. Courtesy of Nature.
“The optical oracle shows that using photonic networks as information carriers opens up unconventional ways to optical computing,” said researcher Cesare Soci, an assistant professor at Nanyang Tech.
He noted that traditional advantages of photonics, such as processing speed, bandwidth and parallelism, could be exploited “in combination with highly reconfigurable materials and nanoscale systems to realize efficient and highly integrated solutions to difficult computational tasks.”
Solving the issue for five nodes was relatively simple, according to the researchers, who added that with this optical oracle, 30 nodes could be solved approximately 375 times faster than with a conventional computer.
This would require a fiber length of 100 to 200 km, which the researchers say is possible with current fiber technology, although it may require optical signal amplification.
"We are currently looking at nonlinear fiber networks, and we are planning to extend this work to integrated photonic networks, which will eventually allow tackling problems of much greater size and complexity,” Soci said.
The optical oracle relies on the length of time it takes a light pulse to travel through the network. A light path experiences a unique delay upon visiting each node. The delays are assigned to the nodes so that their total can be obtained only by summing each node’s delay only once.
The researchers found that if a light pulse traveling through the network is detected after this sum of delay times, it means that a Hamiltonian path does indeed exist in the network; otherwise, such a path does not exist.
This problem is part of a class of famous complexity problems known as NP-complete problems. Conventional computer algorithms can solve such issues when incorporating small numbers, but the time this takes greatly increases with the size of the problem.
Optical telecommunications networks could solve such problems much faster than standard computers. This could have a significant impact on applications such as secure communications, routing optimization and optical data processing, as well as silicon photonics platforms with femtosecond lasers to allow for a more compact architecture.
The researchers also are investigating how these optical networks could be used to mimic the complexities of the human brain.
The research was published in Light: Science & Application
For more information, visit: www.ntu.edu.sg