REACT

In the recent times, there have been myriad of applications ranging from robotics, speech processing and IoT-based devices which use neural network computations. With the increase in machine intelligence, there has been an increase in requirement for specialized hardware at the edge to perform inference and training. This project targets a novel hardware accelerator which can perform inference and training computation in an energy-efficient manner on a small footprint. The accelerator has been designed in a NoC centric fashion which allows it to scale up very effieciently as per the application and device requirements.

This work has been published in DAC 2022.

Please feel free to go through the REACT Project Website