Hardware-aware neural architecture search (HW-Aware NAS) and neural architecture accelerator search (NAAS)

Research Description

Neural architecture search (NAS) provides a way to automatically find a highly accurate NN model without human intervention. However, the conventional NAS approaches come with a major flaw; they focus on increasing only prediction accuracy without considering computational cost of NN inference. This will make NAS impractical in many cases such as execution on mobile devices where computational capacity is limited. Correspondingly, hardware-aware NAS approaches are gaining attention recently as they generate network architectures that can be executed in hardware efficiently maintaining the prediction accuracy.

Neural architecture accelerator search (NAAS) finds both a neural network model and a hardware accelerator at the same time such that they are synergically working together.

Your Job

  • Understanding basic computer architecture.
  • Understanding various neural network architectures.
  • Evaluating how neural architecture affects hardware efficiency.
  • Designing a novel HW-aware NAS or NAAS.

Related Papers:

  1. Accepted
    AlphaAccelerator: An Automatic Neural FPGA Accelerator Design Framework Based on GNNs
    Jiho Lee , Jieui Kang , Eunjin Lee , Yejin Lee , and Jaehyeong Sim
    In 2024 21st International SoC Design Conference (ISOCC)
  2. TD-NAAS: Template-Based Differentiable Neural Architecture Accelerator Search
    HaYoung Lim , Yeseo Jang , Juyeon Kim , and Jaehyeong Sim
    In 2023 20th International SoC Design Conference (ISOCC)