공간적 스파이킹 어텐션 메커니즘을 통한 이벤트 시퀀스 인식

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 15
  • Download : 0
As the necessity for utilizing deep neural networks on low-powered devices increases, interest in Spiking Neural Networks (SNN) which requires very small amount of power is growing. However, due to the non-differential nature of SNN, it is difficult to apply the learning method applied to the existing neural networks, and accordingly, the performance tends to be slightly lower than the conventional neural networks. Therefore, this paper proposes a spatial spiking attention mechanism to improve event data recognition performance of SNN. The spatial spiking attention mechanism applies spatially different weights to the input according to the firing rate of each pixel of the event image. For performance evaluation, a neuromorphic dataset, N-MNIST[1] is used. As a result of the experiment, it was confirmed that the spiking attention mechanism improves recognition performance when it is applied to SNN.
Publisher
제어•로봇•시스템학회
Issue Date
2020-07-02
Language
Korean
Citation

제 35회 제어•로봇•시스템학회 학술대회 (ICROS 2020), pp.462 - 463

URI
http://hdl.handle.net/10203/277818
Appears in Collection
EE-Conference Papers(학술회의논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0