BROWSE
Detail
Conference Paper
A Lightweight on Spiking Generative Adversarial Networks for IoT Applications
Cited 0 time in 
Share 
Share
- Authors
- Seongmo Park, B.G. Choi, Piljae Park, Sungdo Kim, K.W.Park
- Issue Date
- 2023-06
- Citation
- International Technical Conference on Circuits/Systems, Computers and Communications (ITC-CSCC) 2023, pp.907-910
- Publisher
- IEEE
- Language
- English
- Type
- Conference Paper
- Abstract
- The proposed algorithm describes SGAN (Spiking Generative Adversarial Networks) based on hardware for IoT Applications. One way to build an excellent dataset representation is to train Spiking Generative Advisory Networks (SGAN) and networks as sensory data for unsupervised tasks. Training various spiking MNIST datasets showed solid evidence that our spike confrontation pairs learn the repetition of the input data from the sensory data to the scene in both the spike generator and the spike discriminator. This algorithm improved a 20% speed up, 91.4% of accuracy and low power operation of IoT applications compared to other networks. For implementation, it was implemented through RTL coding and logic synthesis based on Xilinx FPGA. Logical synthesis results showed LUT 5%, FF 6%, IO 49%, and BUFG 3%.
- KSP Keywords
- AND logic, IoT Applications, Logic synthesis, Low-power operation, Speed-up, Xilinx FPGA, generative adversarial network, input data, sensory data, spike generator