English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Neural architecture search (NAS) is a promising technique to design efficientand high-performance deep neural networks (DNNs). As the performancerequirements of ML applications grow continuously, the hardware acceleratorsstart playing a central role in DNN design. This trend makes NAS even morecomplicated and time-consuming for most real applications. This paper proposesFLASH, a very fast NAS methodology that co-optimizes the DNN accuracy andperformance on a real hardware platform. As the main theoretical contribution,we first propose the NN-Degree, an analytical metric to quantify thetopological characteristics of DNNs with skip connections (e.g., DenseNets,ResNets, Wide-ResNets, and MobileNets). The newly proposed NN-Degree allows usto do training-free NAS within one second and build an accuracy predictor bytraining as few as 25 samples out of a vast search space with more than 63billion configurations. Second, by performing inference on the target hardware,we fine-tune and validate our analytical models to estimate the latency, area,and energy consumption of various DNN architectures while executing standard MLdatasets. Third, we construct a hierarchical algorithm based on simplicialhomology global optimization (SHGO) to optimize the model-architectureco-design process, while considering the area, latency, and energy consumptionof the target hardware. We demonstrate that, compared to the state-of-the-artNAS approaches, our proposed hierarchical SHGO-based algorithm enables morethan four orders of magnitude speedup (specifically, the execution time of theproposed algorithm is about 0.1 seconds). Finally, our experimental evaluationsshow that FLASH is easily transferable to different hardware architectures,thus enabling us to do NAS on a Raspberry Pi-3B processor in less than 3seconds.

FLASH: Fast Neural Architecture Search with Hardware Optimization