Neural Architecture Search Survey: A Hardware Perspective

We review the problem of automating hardware-aware architectural design process Deep Neural Networks (DNNs). The field Convolutional Network (CNN) algorithm has led to advancements in many fields, such as computer vision, virtual reality, and autonomous driving. end-to-end a CNN is challenging time-consuming task, it requires expertise multiple areas signal image processing, neural networks, optimization. At same time, several hardware platforms, general- special-purpose, have equally contributed training deployment these complex networks different setting. Hardware-Aware Architecture Search (HW-NAS) automates DNNs alleviate human effort generate efficient models accomplishing acceptable accuracy-performance tradeoffs. goal this article provide insights understanding HW-NAS techniques for various platforms (MCU, CPU, GPU, ASIC, FPGA, ReRAM, DSP, VPU), followed by co-search methodologies accelerator specifications.