Photonic integrated circuits, or optical chips, promise a host of advantages over their electronic counterparts, including reduced power consumption and processing speedups.
Writing in its bog it seems that Intel’s boffins have been developing previous work on a type of photonic circuit known as a Mach-Zender interferometer (MZI) which can be configured to perform a two-by-two matrix multiplication between quantities related to the phases of two light beams.
When these small matrix multiplications are arranged in a triangular mesh to create larger matrices, they produce a circuit that implements a matrix-vector multiplication, a core computation in deep learning.
Chipzilla looked at two architectures for building an AI system out of MZIs: GridNet and FFTNet.
GridNet predictably arranges the MZIs in a grid, while FFTNet slots them into a butterfly-like pattern. After training the two in simulation on a benchmark deep learning task of handwritten digit recognition (MNIST), the researchers found that GridNet achieved higher accuracy than FFTNet (98 per cent versus 95 per cent) in the case of double-precision floating point accuracy, but that FFTNet was “significantly more robust.” In fact, GridNet’s performance fell below 50 per cent with the addition of artificial noise, while FFTNet’s remained nearly constant.
All this apparently lays the groundwork for AI software training techniques that might obviate the need to fine-tune optical chips post-manufacturing, saving valuable time and labour.
Intel AI products group senior director Casimir Wierzynski wrote that there are usually imperfections in any manufacturing process which means that there will be small variations within and across chips, and these will affect the accuracy of computations.
“If ONNs are to become a viable piece of the AI hardware ecosystem, they will need to scale up to larger circuits and industrial manufacturing techniques … Our results suggest that choosing the right architecture in advance can greatly increase the probability that the resulting circuits will achieve their desired performance even in the face of manufacturing variations.”