A new support vector machine with an optimal additive kernel

Jeonghyun Baek, Euntai Kim

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Although the kernel support vector machine (SVM) outperforms linear SVM, its application to real world problems is limited because the evaluation of its decision function is computationally very expensive due to kernel expansion. On the other hand, additive kernel (AK) SVM enables fast evaluation of a decision function using look-up tables (LUTs). The AKs, however, assume a specific functional form for kernels such as the intersection kernel (IK) or χ 2 kernel, and are problematic in that their performance is seriously degraded when a given problem is highly nonlinear. To address this issue, an optimal additive kernel (OAK) is proposed in this paper. The OAK does not assume any specific kernel form, but the kernel is represented by a quantized Gram table. The training of the OAK SVM is formulated as semi-definite programming (SDP), and it is solved by convex optimization. In the experiment, the proposed method is tested with 2D synthetic datasets, UCI repository datasets and LIBSVM datasets. The experimental results show that the proposed OAK SVM has better performance than the previous AKs and RBF kernel while maintaining fast computation using look-up tables.

Original languageEnglish
Pages (from-to)279-299
Number of pages21
JournalNeurocomputing
Volume329
DOIs
Publication statusPublished - 2019 Feb 15

Fingerprint

Support vector machines
Convex optimization
Support Vector Machine
Datasets
Experiments

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Cognitive Neuroscience
  • Artificial Intelligence

Cite this

@article{bebed7dec43c4bdc80faeb8211499043,
title = "A new support vector machine with an optimal additive kernel",
abstract = "Although the kernel support vector machine (SVM) outperforms linear SVM, its application to real world problems is limited because the evaluation of its decision function is computationally very expensive due to kernel expansion. On the other hand, additive kernel (AK) SVM enables fast evaluation of a decision function using look-up tables (LUTs). The AKs, however, assume a specific functional form for kernels such as the intersection kernel (IK) or χ 2 kernel, and are problematic in that their performance is seriously degraded when a given problem is highly nonlinear. To address this issue, an optimal additive kernel (OAK) is proposed in this paper. The OAK does not assume any specific kernel form, but the kernel is represented by a quantized Gram table. The training of the OAK SVM is formulated as semi-definite programming (SDP), and it is solved by convex optimization. In the experiment, the proposed method is tested with 2D synthetic datasets, UCI repository datasets and LIBSVM datasets. The experimental results show that the proposed OAK SVM has better performance than the previous AKs and RBF kernel while maintaining fast computation using look-up tables.",
author = "Jeonghyun Baek and Euntai Kim",
year = "2019",
month = "2",
day = "15",
doi = "10.1016/j.neucom.2018.10.032",
language = "English",
volume = "329",
pages = "279--299",
journal = "Review of Economic Dynamics",
issn = "1094-2025",
publisher = "Academic Press Inc.",

}

A new support vector machine with an optimal additive kernel. / Baek, Jeonghyun; Kim, Euntai.

In: Neurocomputing, Vol. 329, 15.02.2019, p. 279-299.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A new support vector machine with an optimal additive kernel

AU - Baek, Jeonghyun

AU - Kim, Euntai

PY - 2019/2/15

Y1 - 2019/2/15

N2 - Although the kernel support vector machine (SVM) outperforms linear SVM, its application to real world problems is limited because the evaluation of its decision function is computationally very expensive due to kernel expansion. On the other hand, additive kernel (AK) SVM enables fast evaluation of a decision function using look-up tables (LUTs). The AKs, however, assume a specific functional form for kernels such as the intersection kernel (IK) or χ 2 kernel, and are problematic in that their performance is seriously degraded when a given problem is highly nonlinear. To address this issue, an optimal additive kernel (OAK) is proposed in this paper. The OAK does not assume any specific kernel form, but the kernel is represented by a quantized Gram table. The training of the OAK SVM is formulated as semi-definite programming (SDP), and it is solved by convex optimization. In the experiment, the proposed method is tested with 2D synthetic datasets, UCI repository datasets and LIBSVM datasets. The experimental results show that the proposed OAK SVM has better performance than the previous AKs and RBF kernel while maintaining fast computation using look-up tables.

AB - Although the kernel support vector machine (SVM) outperforms linear SVM, its application to real world problems is limited because the evaluation of its decision function is computationally very expensive due to kernel expansion. On the other hand, additive kernel (AK) SVM enables fast evaluation of a decision function using look-up tables (LUTs). The AKs, however, assume a specific functional form for kernels such as the intersection kernel (IK) or χ 2 kernel, and are problematic in that their performance is seriously degraded when a given problem is highly nonlinear. To address this issue, an optimal additive kernel (OAK) is proposed in this paper. The OAK does not assume any specific kernel form, but the kernel is represented by a quantized Gram table. The training of the OAK SVM is formulated as semi-definite programming (SDP), and it is solved by convex optimization. In the experiment, the proposed method is tested with 2D synthetic datasets, UCI repository datasets and LIBSVM datasets. The experimental results show that the proposed OAK SVM has better performance than the previous AKs and RBF kernel while maintaining fast computation using look-up tables.

UR - http://www.scopus.com/inward/record.url?scp=85056328007&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85056328007&partnerID=8YFLogxK

U2 - 10.1016/j.neucom.2018.10.032

DO - 10.1016/j.neucom.2018.10.032

M3 - Article

AN - SCOPUS:85056328007

VL - 329

SP - 279

EP - 299

JO - Review of Economic Dynamics

JF - Review of Economic Dynamics

SN - 1094-2025

ER -