Neuro-Photonix: Enabling Near-Sensor Neuro-Symbolic AI Computing on\n Silicon Photonics Substrate

Neuro-symbolic Artificial Intelligence (AI) models, blending neural networks\nwith symbolic AI, have facilitated transparent reasoning and context\nunderstanding without the need for explicit rule-based programming. However,\nimplementing such models in the Internet of Things (IoT) sensor nodes presents\nhurdles due to computational constraints and intricacies. In this work, for the\nfirst time, we propose a near-sensor neuro-symbolic AI computing accelerator\nnamed Neuro-Photonix for vision applications. Neuro-photonix processes neural\ndynamic computations on analog data while inherently supporting\ngranularity-controllable convolution operations through the efficient use of\nphotonic devices. Additionally, the creation of an innovative, low-cost ADC\nthat works seamlessly with photonic technology removes the necessity for costly\nADCs. Moreover, Neuro-Photonix facilitates the generation of HyperDimensional\n(HD) vectors for HD-based symbolic AI computing. This approach allows the\nproposed design to substantially diminish the energy consumption and latency of\nconversion, transmission, and processing within the established cloud-centric\narchitecture and recently designed accelerators. Our device-to-architecture\nresults show that Neuro-Photonix achieves 30 GOPS/W and reduces power\nconsumption by a factor of 20.8 and 4.1 on average on neural dynamics compared\nto ASIC baselines and photonic accelerators while preserving accuracy.\n