New self-configuring optical chip performs various functions

Researchers from the Huazhong University of Science and Technology in China have developed an optical chip that can configure itself to achieve various functions. The positive real-value matrix computation they have achieved gives the chip the potential to be used in applications requiring optical neural networks. Optical neural networks are suitable for data-heavy tasks such as image classification, gesture interpretation and speech recognition.

Photonic integrated circuits that can be reconfigured after manufacturing to perform different functions have been developed before; however, they can be difficult to configure because the user must understand the internal structure and principles of the chip and individually adjust its basic units. Research team leader Jianji Dong said the optical chip can be treated as a black box, so users don’t need to understand its internal structure to change its function. “They only need to set a training objective and, with computer control, the chip will self-configure to achieve the desired functionality based on the input and output,” Dong said.

The new optical chip is based on a network of waveguide-based optical components called Mach–Zehnder interferometers (MZIs) arranged in a quadrilateral pattern. The researchers showed that the chip can self-configure to perform optical routing, low-loss light energy splitting and the matrix computations used to create neural networks.

“In the future, we anticipate the realisation of larger-scale on-chip programmable waveguide networks. With additional development, it may become possible to achieve optical functions comparable to those of field-programmable gate arrays (FPGAs) — electrical integrated circuits that can be reprogrammed to perform any desired application after they are manufactured,” Dong said.

The on-chip quadrilateral MZI network could be useful for applications involving optical neural networks, which are created from networks of interconnected nodes. To use an optical neural network, the network must be trained with known data to determine the weights between each pair of nodes — a process that requires matrix multiplication. According to Dong, on-chip matrix operations are typically implemented using forward-propagating MZI networks or microring arrays. “Inspired by FPGAs in electronics, we wanted to use an MZI topological network structure that allowed both feedforward and feedbackward propagation for matrix operations,” Dong said.

The new optical chip can be reconfigured by adjusting the voltages of electrodes, which creates light propagation paths in the quadrilateral network. The researchers integrated a gradient descent algorithm to accelerate the convergence rate of the cost function, which gauges the accuracy of the network with each training iteration. After each training iteration, the chip updates the voltages of all the adjustable electrodes — rather than the value of a single variable — which improves the convergence rate of cost function. These improvements help make the training process faster.

The chip can be used to perform positive real matrix computation, with the researchers verifying the feasibility of this in a quadrilateral MZI network. The error between the chip’s training results and the target matrices was minimal. The researchers also demonstrated optical routing — a specialised case of positive real matrix computation — with a high extinction ratio. Optical routing can route optical signals between equipment, such as processors and memory units, in data centres.

The researchers are now working to make improvements to the chip that would allow even more matrix operation capabilities. They would also like to explore using it for other applications of matrix computing beyond optical neural networks.

The research findings were published in the journal Optical Materials Express.

Image credit: iStock.com/da-kuk