An Energy-Efficient Silicon Photonic-Assisted Deep Learning Accelerator for Big Data

Abstract

Deep learning has become the most mainstream technology in artificial intelligence (AI) because it can be comparable to human performance in complex tasks. However, in the era of big data, the ever-increasing data volume and model scale makes deep learning require mighty computing power and acceptable energy costs. For electrical chips, including most deep learning accelerators, transistor performance limitations make it challenging to meet computing’s energy efficiency requirements. Silicon photonic devices are expected to replace transistors and become the mainstream components in computing architecture due to their advantages, such as low energy consumption, large bandwidth, and high speed. Therefore, we propose a silicon photonic-assisted deep learning accelerator for big data. The accelerator uses microring resonators (MRs) to form a photonic multiplication array. It combines photonic-specific wavelength division multiplexing (WDM) technology to achieve multiple parallel calculations of input feature maps and convolution kernels at the speed of light, providing the promise of energy efficiency and calculation speed improvement. The proposed accelerator achieves at least a 75x improvement in computational efficiency compared to the traditional electrical design.

An Improved Receptive Fields Network for Matching Remote Sensing Images

“…The effect of key point detection is similar to the SIFT algorithm with good robustness, but it cannot accurately extract the feature points of smooth edges. Unlike the LIFT method, SuperPoint [5] uses self-supervised training feature point detection and feature descriptor extraction, the feature is extracted through VGG [6], the key point detection part needs to be pre-trained on the synthetic image data set, and the entire network needs to be trained on the synthetically transformed image. LF-Net [2] uses the Siamense structure [7] without the need of any manual method, it generates a feature map through the deep feature extraction network, which can extract deep features of larger receptive field from the input image, but the shallow features will be lost.…”