2018 Low-Power Image Recognition Challenge

Alyamkin, Sergei; Ardi, Matthew; Brighton, Achille; Berg, Alexander C.; Chen, Yiran; Cheng, Hsin-Pai; Chen, Bin; Zhang, Fan; Chen, Feng; Fu, Bo; Gauen, Kent; Go, Jongkook; Goncharenko, Alexander A.; Guo, Xunhua; Nguyen, Hong Hanh; Howard, Andrew W.; Huang, Yue-Xin; Kang, Dong‐Hyun; Kim, Jaeyoun; Kondratyev, Alexander S.; Lee, Seungjae; Lee, Suwoong; Lee, Junhyeok; Liang, Zhiyu; Liu, Xin; Liu, Juzheng; Li, Zichao; Yang, Ling; Lü, Yanbin; Malik, Deeptanshu; Park, Eunbyung; Repin, Denis; Sheng, Tao; Shen, Liang; Sun, Fengrui; Svitov, David; Thiruvathukal, George K.; Zhang, Baiwu; Zhang, Jingchi; Zhang, Xiaopeng; Zhuo, Shaojie

doi:10.48550/arxiv.1810.01732

Cited by 2 publications

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Note also that although the number of MAC operations does not directly translate to the CONV layer runtime [33], it Fig. 5 The 3D index space of a matrix-by-matrix multiply-add (a) and its corresponding linear projections onto a 2D array processor space along R 2 -edge (b), M-edge (c), and…”

Section: Direct Vector-parallel Computingmentioning

confidence: 99%

In Search of the Performance- and Energy-Efficient CNN Accelerators

Sedukhin

Yamamoto

2022

IEICE Trans. Electron.

View full text Add to dashboard Cite

In this paper, starting from the algorithm, a performanceand energy-efficient 3D structure or shape of the Tensor Processing Engine (TPE) for CNN acceleration is systematically searched and evaluated. An optimal accelerator's shape maximizes the number of concurrent MAC operations per clock cycle while minimizes the number of redundant operations. The proposed 3D vector-parallel TPE architecture with an optimal shape can be very efficiently used for considerable CNN acceleration. Due to implemented support of inter-block image data independency, it is possible to use multiple of such TPEs for the additional CNN acceleration. Moreover, it is shown that the proposed TPE can also be uniformly used for acceleration of the different CNN models such as VGG, ResNet, YOLO, and SSD. We also demonstrate that our theoretical efficiency analysis is matched with the result of a real implementation for an SSD model to which a state-of-the-art channel pruning technique is applied.

show abstract

Section: Direct Vector-parallel Computingmentioning

confidence: 99%