TY - GEN
T1 - Distributed Parallel Deep Learning for Fast Extraction of Similar Weather Map
AU - Kang, Boseon
AU - Jeong, Jae Heon
AU - Jeong, Changsung
N1 - Funding Information:
ACKNOWLEDGMENT This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2017R1D1A1B03035461), the Brain Korea 21 Plus Project in 2018, and the Institute for Information & communications Technology Promotion(IITP) grant funded by the Korean government (MSIP) (No. 2018-0-00739, Deep learning-based natural language contents evaluation technology for detecting fake news) and the utilization of satellite information project through the Korea Aerospace Research Institute (KARI).
Publisher Copyright:
© 2018 IEEE.
PY - 2019/2/22
Y1 - 2019/2/22
N2 - For real time weather forecasting, it is necessary to search most similar weather map very fast among a large amount of data accumulated so far. Recently, deep learning is used for more accurate weather forecasting. However, it takes a huge amount of time for training deep learning model in order to process a number of previous weather maps. In this paper, we shall present fast distributed parallel algorithms for training deep neural network model based on CNN on parallel and distributed environment with GPUs for various number of models in order to extract most similar weather map from CNN. For each case of single and multi nodes, we compare the performance of our algorithm increasing the number of GPUs, and for the case of multi nodes, compare the performance for two ways of communications: synchronous and asynchronous. Also, we shall show the performance of our algorithm for the various number of models on single and multi nodes.
AB - For real time weather forecasting, it is necessary to search most similar weather map very fast among a large amount of data accumulated so far. Recently, deep learning is used for more accurate weather forecasting. However, it takes a huge amount of time for training deep learning model in order to process a number of previous weather maps. In this paper, we shall present fast distributed parallel algorithms for training deep neural network model based on CNN on parallel and distributed environment with GPUs for various number of models in order to extract most similar weather map from CNN. For each case of single and multi nodes, we compare the performance of our algorithm increasing the number of GPUs, and for the case of multi nodes, compare the performance for two ways of communications: synchronous and asynchronous. Also, we shall show the performance of our algorithm for the various number of models on single and multi nodes.
KW - Deep learning
KW - Distributed Environment
KW - Similar weather map
KW - Weather Prediction
UR - http://www.scopus.com/inward/record.url?scp=85063199628&partnerID=8YFLogxK
U2 - 10.1109/TENCON.2018.8650104
DO - 10.1109/TENCON.2018.8650104
M3 - Conference contribution
AN - SCOPUS:85063199628
T3 - IEEE Region 10 Annual International Conference, Proceedings/TENCON
SP - 1426
EP - 1429
BT - Proceedings of TENCON 2018 - 2018 IEEE Region 10 Conference
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2018 IEEE Region 10 Conference, TENCON 2018
Y2 - 28 October 2018 through 31 October 2018
ER -