Training Large-Scale Optical Neural Networks with Two-Pass Forward Propagation
This work addresses training bottlenecks for optical computing systems, advancing their potential for complex data tasks, though it appears incremental in the broader context of neural network methods.
The paper tackles training efficiency, nonlinear function implementation, and large input data processing in Optical Neural Networks by introducing Two-Pass Forward Propagation, a novel training method that avoids specific nonlinear activations and includes a new convolutional implementation, resulting in significant improvements in training speed, energy efficiency, and scalability.
This paper addresses the limitations in Optical Neural Networks (ONNs) related to training efficiency, nonlinear function implementation, and large input data processing. We introduce Two-Pass Forward Propagation, a novel training method that avoids specific nonlinear activation functions by modulating and re-entering error with random noise. Additionally, we propose a new way to implement convolutional neural networks using simple neural networks in integrated optical systems. Theoretical foundations and numerical results demonstrate significant improvements in training speed, energy efficiency, and scalability, advancing the potential of optical computing for complex data tasks.