Towards a Design Framework for TNN-Based Neuromorphic Sensory Processing Units

TL;DR

This paper discusses developing a design framework for application-specific TNN-based neuromorphic sensory processing units, aiming to improve energy efficiency and facilitate rapid hardware-software co-design for sensory tasks.

Contribution

It introduces a custom design framework and tools for TNN-based NSPUs, enabling efficient exploration and optimization of application-specific neuromorphic sensory processors.

Findings

Reviewed existing NSPU designs for time-series and image tasks.

Proposed a software-hardware co-design flow using EDA tools.

Outlined future research directions for NSPU development.

Abstract

Temporal Neural Networks (TNNs) are spiking neural networks that exhibit brain-like sensory processing with high energy efficiency. This work presents the ongoing research towards developing a custom design framework for designing efficient application-specific TNN-based Neuromorphic Sensory Processing Units (NSPUs). This paper examines previous works on NSPU designs for UCR time-series clustering and MNIST image classification applications. Current ideas for a custom design framework and tools that enable efficient software-to-hardware design flow for rapid design space exploration of application-specific NSPUs while leveraging EDA tools to obtain post-layout netlist and power-performance-area (PPA) metrics are described. Future research directions are also outlined.