TY - GEN
T1 - Particle swarm stepwise algorithm (PaSS) on multicore hybrid CPU-GPU clusters
AU - Yang, Mu
AU - Chen, Ray Bing
AU - Chung, I. Hsin
AU - Wang, Weichung
N1 - Publisher Copyright:
© 2016 IEEE.
PY - 2017/3/10
Y1 - 2017/3/10
N2 - Variable (feature) selection is a key component in artificial intelligence. One way to perform variable selection is to solve the information-criterion-based optimization problems. These optimization problems arise from data mining, genomes analysis, machine learning, numerical simulations, and others. Particle Swarm Stepwise Algorithm (PaSS) is a stochastic search algorithm proposed to solve the information-criterion-based variable selection optimization problems. It has been shown recently that the PaSS outperforms several existed methods. However, to solve the target optimization problems remains a challenge due to the large search spaces. We tackle this issue by proposing a parallel version of the PaSS on clusters equipped with CPU and GPU to shorten the computational time without compromise in solution accuracy. We have successfully achieved near-linear scalability on CPU with single to 64 threads and gained further 7X faster timing performance by using GPU.
AB - Variable (feature) selection is a key component in artificial intelligence. One way to perform variable selection is to solve the information-criterion-based optimization problems. These optimization problems arise from data mining, genomes analysis, machine learning, numerical simulations, and others. Particle Swarm Stepwise Algorithm (PaSS) is a stochastic search algorithm proposed to solve the information-criterion-based variable selection optimization problems. It has been shown recently that the PaSS outperforms several existed methods. However, to solve the target optimization problems remains a challenge due to the large search spaces. We tackle this issue by proposing a parallel version of the PaSS on clusters equipped with CPU and GPU to shorten the computational time without compromise in solution accuracy. We have successfully achieved near-linear scalability on CPU with single to 64 threads and gained further 7X faster timing performance by using GPU.
UR - https://www.scopus.com/pages/publications/85017331494
UR - https://www.scopus.com/pages/publications/85017331494#tab=citedBy
U2 - 10.1109/CIT.2016.101
DO - 10.1109/CIT.2016.101
M3 - Conference contribution
AN - SCOPUS:85017331494
T3 - Proceedings - 2016 16th IEEE International Conference on Computer and Information Technology, CIT 2016, 2016 6th International Symposium on Cloud and Service Computing, IEEE SC2 2016 and 2016 International Symposium on Security and Privacy in Social Networks and Big Data, SocialSec 2016
SP - 265
EP - 272
BT - Proceedings - 2016 16th IEEE International Conference on Computer and Information Technology, CIT 2016, 2016 6th International Symposium on Cloud and Service Computing, IEEE SC2 2016 and 2016 International Symposium on Security and Privacy in Social Networks and Big Data, SocialSec 2016
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 16th IEEE International Conference on Computer and Information Technology, CIT 2016
Y2 - 7 December 2016 through 10 December 2016
ER -