Digital Oscillatory Neural Networks for AI Edge Applications
Abstract
This work focuses on Oscillatory Neural Network (ONN) computing paradigm for edge artificial intelligence applications. In particular, we use a digital ONN implemented on FPGA to explore novel architectures, learning algorithms, and applications. First, using a fully-connected ONN architecture, ONN can perform pattern recognition, which can be applied for various edge applications, like image processing and robotics. Then, this work also introduces layered ONN architectures applied to image edge detection and feature detection.