A team of researchers from Stanford University have got success in demonstrating that artificial neural networks (ANNs) can be trained directly on an optical chip. With their latest work, the researchers showed that an optical circuit can perform a critical function of an electronics-based ANN.
According to researchers, their work could help find out faster and more energy-efficient methods to solve complex problems and carry out tasks such as image or speech recognition.
Shanhui Fan, the lead researcher from Stanford University said their results could help improve the capability of ANNs to perform tasks required for self-driving cars.
In an ANN, information is processed through connected units. This information processing mechanism is similar to mechanism used by the brain to processes information.
For using ANNs to perform a complex task, researchers first need to train algorithms to categorize inputs.
In their paper, researchers discuss how optical neural networks can be trained directly in the device by implementing an optical analogue of the ‘backpropagation’ algorithm.
“Using a physical device rather than a computer model for training makes the process more accurate,” said Tyler W. Hughes, first author of the paper.
“Also, because the training step is a very computationally expensive part of the implementation of the neural network, performing this step optically is the key to improving the computational efficiency, speed and power consumption of artificial networks.”
The researchers were able to implement an all-optical neural network by designing an optical chip that replicates the way that conventional computers train neural networks.
According to Hughes, their approach speeds up training significantly, especially for large networks.
The researchers now want to further optimize the system and use it to implement a practical application of a neural network task.
The detailed finding of the study titled “Training of photonic neural networks through in situ backpropagation and gradient measurement” were published in Optica, The Optical Society’s journal for high impact research.