Optical Chip as a Neural Network

Researchers at the National Institute of Standards and Technology NIST have made a silicon chip that distributes optical signals precisely across a miniature brain-like grid, showcasing a potential new design for neural networks. The team proposes to use light instead of electricity as a signaling medium. Neural networks already have demonstrated remarkable power in solving complex problems, including rapid pattern recognition and data analysis. The use of light would eliminate interference due to electrical charge and the signals would travel faster and farther.

This grid-on-a-chip distributes light signals precisely, showcasing a potential new design for neural networks. The three-dimensional structure enables complex routing scheme, which are necessary to mimic the brain. (Source: Chiles, NIST)

“Light’s advan­tages could improve the perfor­mance of neural nets for scientific data analysis such as searches for Earth-like planets and quantum infor­mation science, and accelerate the develop­ment of highly intuitive control systems for auto­nomous vehicles,” NIST physicist Jeff Chiles said. A conven­tional computer processes information through algo­rithms, or human-coded rules. By contrast, a neural network relies on a network of connec­tions among processing elements, or neurons, which can be trained to recog­nize certain patterns of stimuli. A neural or neuro­morphic computer would consist of a large, complex system of neural networks.

The new chip overcomes a major challenge to the use of light signals by verti­cally stacking two layers of photonic wave­guides. This three-dimen­sional design enables complex routing schemes, which are necessary to mimic neural systems. Further­more, this design can easily be extended to incor­porate addi­tional wave­guiding layers when needed for more complex networks. The stacked wave­guides form a three-dimen­sional grid with 10 inputs or upstream neurons each connecting to 10 outputs or downstream neurons, for a total of 100 receivers. Fabri­cated on a silicon wafer, the wave­guides are made of silicon nitride and are each 800 nanometers wide and 400 nm thick. Researchers created software to auto­matically generate signal routing, with adjustable levels of connec­tivity between the neurons.

Laser light was directed into the chip through an optical fiber. The goal was to route each input to every output group, following a selected distri­bution pattern for light intensity or power. Power levels represent the pattern and degree of connec­tivity in the circuit. The researchers demons­trated two schemes for control­ling output inten­sity: uniform – each output receives the same power, and a bell curve distri­bution in which middle neurons receive the most power, while peri­pheral neurons receive less.

To evaluate the results, researchers made images of the output signals. All signals were focused through a micro­scope lens onto a semicon­ductor sensor and processed into image frames. This method allows many devices to be analyzed at the same time with high precision. The output was highly uniform, with low error rates, confir­ming precise power distri­bution. “We’ve really done two things here,” Chiles said. “We’ve begun to use the third dimension to enable more optical connec­tivity, and we’ve developed a new measure­ment technique to rapidly charac­terize many devices in a photonic system. Both advances are crucial as we begin to scale up to massive opto­electronic neural systems.” (Source: NIST)

Reference: J. Chiles et al.: Design, fabrication, and metrology of 10 x 100 multi-planar integrated photonic routing manifolds for neural networks, APL Phot. 3, 106101 (2018); DOI: 10.1063/1.5039641

Link: Applied Physics Division, National Institute of Standards and Technology, Boulder, USA

Speak Your Mind

*