http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Slightly-slacked dropout for improving neural network learning on FPGA
Sota Sawaguchi,Hiroaki Nishi 한국통신학회 2018 ICT Express Vol.4 No.2
Neural Network Learning (NNL) is compute-intensive. It often involves a dropout technique which effectively regularizes the network to avoid overfitting. As such, a hardware accelerator for dropout NNL has been proposed; however, the existing method encounters a huge transfer cost between hardware and software. This paper proposes Slightly-Slacked Dropout (SS-Dropout), a novel deterministic dropout technique to address the transfer cost while accelerating the process. Experimental results show that our SS-Dropout technique improves both the usual and dropout NNL accelerator, i.e., 1.55 times speed-up and three order-of-magnitude less transfer cost, respectively.
A privacy-preserving sharing method of electricity usage using self-organizing map
Yuichi Nakamura,Keiya Harada,Hiroaki Nishi 한국통신학회 2018 ICT Express Vol.4 No.1
Smart meters for measuring electricity usage are expected in electricity usage management. Although the relevant power supplier stores the measured data, the data are worth sharing among power suppliers because the entire data of a city will be required to control the regional grid stability or demand–supply balance. Even though many techniques and methods of privacy-preserving data mining have been studied to share data while preserving data privacy, a study on sharing electricity usage data is still lacking. In this paper, we propose a sharing method of electricity usage while preserving data privacy using a self-organizing map.