A Reconfigurable Photonic Processor for NP-complete Problems

TL;DR

This paper introduces a reconfigurable photonic processor capable of efficiently solving NP-complete problems like the subset sum problem, leveraging high-speed photon propagation and parallelism to outperform traditional electronic processors.

Contribution

It presents a novel reconfigurable photonic architecture that can adapt to different problem instances and potentially solve large-scale NP-complete problems more efficiently.

Findings

Photonic processor surpasses recent electronic processors in solving SSP.

The architecture is fully reconfigurable with up to 2^N configurations.

Experimental results demonstrate the potential for large-scale hard problem solving.

Abstract

NP-complete problems are widely and deeply involved in various real-life scenarios while still intractable to solve efficiently on conventional computers. It is of great practical significance to construct versatile computing architectures that solve NP-complete problems with computational advantage. Here, we present a reconfigurable photonic processor to efficiently solve a benchmark NP-complete problem, the subset sum problem (SSP). We show that in the case of successive primes, the photonic processor has genuinely surpassed commercial electronic processors launched recently by taking advantages of the high propagation speed and vast parallelism of photons and state-of-the-art integrated photonic technology. Moreover, we are able to program the photonic processor to tackle different problem instances relying on the tunable integrated modules, variable split junctions, which can be used to build a fully reconfigurable architecture potentially allowing $2^{N}$ configurations at most. Our experiments confirm the potential of the photonic processor as a versatile and efficient computing platform, suggesting a possible practical route to solving computationally hard problems at large scale.