Neuromorphic processing at 11 Tera-OPs with soliton crystal Kerr microcombs
Mengxi Tan, Xingyuan Xu, Jiayang Wu, Andreas Boes, Bill Corcoran, Thach Giang Nguyen, S. T. Chu, Brent E. Little, Roberto Morandotti, Arnan Mitchell, David Moss · 2021
We report a new approach to ONNs based on integrated Kerr micro-combs that is programmable, highly scalable and capable of reaching ultra-high speeds. We demonstrate a single neuron perceptron at 11.9 Giga-OPS at 8 bits per OP, or 95.2 Gbps. We then demonstrate a convolutional accelerator operating beyond 11 TeraOPs/s (TOPs). We test the perceptron on handwritten-digit recognition and cancer-cell detection - achieving over 90% and 85% accuracy, respectively.