TY - JOUR
T1 - PLAME
T2 - Piecewise-Linear Approximate Measure for Additive Kernel SVM
AU - Chan, Tsz Nam
AU - Li, Zhe
AU - Hou, L.
AU - Cheng, Reynold
N1 - Funding Information:
This work was supported in part by the NSFC under Grant 62202401, in part by the Science and Technology Development Fund Macau under Grants SAR 0015/2019/AKP, 0031/2022/A, SKL-IOTSC-2021-2023, in part by the Research Grant of University of Macau under Grant MYRG2022- 00252-FST, in part by Wuyi University Hong Kong and Macau joint Research Fund under Grant 2021WGALH14, in part by the University of Hong Kong under Grants 104005858 and 10400599, and in part by the Guangdong-Hong Kong-Macau Joint Laboratory Program 2020 under Grant 2020B1212030009.
Publisher Copyright:
© 2023 IEEE.
PY - 2023/10/1
Y1 - 2023/10/1
N2 - Additive Kernel SVM has been extensively used in many applications, including human activity detection and pedestrian detection. Since training an additive kernel SVM model is very time-consuming, which is not scalable to large-scale datasets, many efficient solutions have been developed in the past few years. However, most of the existing methods normally fail to achieve one of these three important conditions which are (1) low classification error, (2) low memory space, and (3) low training time. In order to simultaneously fulfill these three conditions, we develop the new piecewise-linear approximate measure (PLAME) for additive kernels. By incorporating PLAME with the well-known dual coordinate descent method, we theoretically show that this approach can achieve the above three conditions. Experimental results on twelve real datasets show that our approach can achieve the best trade-off between the accuracy, memory space, and training time compared with different types of state-of-the-art methods.
AB - Additive Kernel SVM has been extensively used in many applications, including human activity detection and pedestrian detection. Since training an additive kernel SVM model is very time-consuming, which is not scalable to large-scale datasets, many efficient solutions have been developed in the past few years. However, most of the existing methods normally fail to achieve one of these three important conditions which are (1) low classification error, (2) low memory space, and (3) low training time. In order to simultaneously fulfill these three conditions, we develop the new piecewise-linear approximate measure (PLAME) for additive kernels. By incorporating PLAME with the well-known dual coordinate descent method, we theoretically show that this approach can achieve the above three conditions. Experimental results on twelve real datasets show that our approach can achieve the best trade-off between the accuracy, memory space, and training time compared with different types of state-of-the-art methods.
KW - Additive kernels
KW - PLAME
KW - SVM
UR - http://www.scopus.com/inward/record.url?scp=85149891531&partnerID=8YFLogxK
U2 - 10.1109/TKDE.2023.3253263
DO - 10.1109/TKDE.2023.3253263
M3 - Journal article
AN - SCOPUS:85149891531
SN - 1041-4347
VL - 35
SP - 9985
EP - 9997
JO - IEEE Transactions on Knowledge and Data Engineering
JF - IEEE Transactions on Knowledge and Data Engineering
IS - 10
ER -