Optimizing kernel functions using transfer learning from unlabeled data

M. Ehsan Abbasnejad; Dhanesh Ramachandram; Rajeswari Mandava

doi:10.1109/ICMV.2009.10

Optimizing kernel functions using transfer learning from unlabeled data

M. Ehsan Abbasnejad
, Dhanesh Ramachandram
, Rajeswari Mandava

Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › Research › peer-review

4 Citations (Scopus)

Abstract

In this paper, we propose an approach to learn the kernel which uses transferred knowledge from unlabeled data to cope with situations where training examples are scarce. In our approach, unlabeled data has been used to construct an optimized kernel that better generalizes on the target dataset. For the proposed kernel learning algorithm, Fisher Discriminant Analysis (FDA) is used in conjunction with Maximum Mean Discrepancy (MMD) test of statistics to optimize a base kernel using labeled and unlabeled data. Thereafter, the constructed kernel from both labeled and unlabeled datasets is used in SVM to evaluate the results which proved to increase prediction accuracy.

Original language	English
Title of host publication	2009 2nd International Conference on Machine Vision, ICMV 2009
Pages	111-117
Number of pages	7
DOIs	https://doi.org/10.1109/ICMV.2009.10
Publication status	Published - 2009
Externally published	Yes
Event	International Conference on Machine Vision 2009 - Dubai, United Arab Emirates Duration: 28 Dec 2009 → 30 Dec 2009 Conference number: 2nd https://ieeexplore.ieee.org/xpl/conhome/5379704/proceeding (Proceedings)

Conference

Conference	International Conference on Machine Vision 2009
Abbreviated title	ICMV 2009
Country/Territory	United Arab Emirates
City	Dubai
Period	28/12/09 → 30/12/09
Internet address	https://ieeexplore.ieee.org/xpl/conhome/5379704/proceeding (Proceedings)

Keywords

Kernel methods
Learning the kernels
Machine learning
Support Vector Machine (SVM)
Transfer learning

Access to Document

10.1109/ICMV.2009.10

Cite this

@inproceedings{db236210fab94f19a0f7b62ef230b045,

title = "Optimizing kernel functions using transfer learning from unlabeled data",

abstract = "In this paper, we propose an approach to learn the kernel which uses transferred knowledge from unlabeled data to cope with situations where training examples are scarce. In our approach, unlabeled data has been used to construct an optimized kernel that better generalizes on the target dataset. For the proposed kernel learning algorithm, Fisher Discriminant Analysis (FDA) is used in conjunction with Maximum Mean Discrepancy (MMD) test of statistics to optimize a base kernel using labeled and unlabeled data. Thereafter, the constructed kernel from both labeled and unlabeled datasets is used in SVM to evaluate the results which proved to increase prediction accuracy.",

keywords = "Kernel methods, Learning the kernels, Machine learning, Support Vector Machine (SVM), Transfer learning",

author = "Abbasnejad, \{M. Ehsan\} and Dhanesh Ramachandram and Rajeswari Mandava",

year = "2009",

doi = "10.1109/ICMV.2009.10",

language = "English",

isbn = "9780769539447",

pages = "111--117",

booktitle = "2009 2nd International Conference on Machine Vision, ICMV 2009",

note = "International Conference on Machine Vision 2009, ICMV 2009 ; Conference date: 28-12-2009 Through 30-12-2009",

url = "https://ieeexplore.ieee.org/xpl/conhome/5379704/proceeding",

}

TY - GEN

T1 - Optimizing kernel functions using transfer learning from unlabeled data

AU - Abbasnejad, M. Ehsan

AU - Ramachandram, Dhanesh

AU - Mandava, Rajeswari

N1 - Conference code: 2nd

PY - 2009

Y1 - 2009

N2 - In this paper, we propose an approach to learn the kernel which uses transferred knowledge from unlabeled data to cope with situations where training examples are scarce. In our approach, unlabeled data has been used to construct an optimized kernel that better generalizes on the target dataset. For the proposed kernel learning algorithm, Fisher Discriminant Analysis (FDA) is used in conjunction with Maximum Mean Discrepancy (MMD) test of statistics to optimize a base kernel using labeled and unlabeled data. Thereafter, the constructed kernel from both labeled and unlabeled datasets is used in SVM to evaluate the results which proved to increase prediction accuracy.

AB - In this paper, we propose an approach to learn the kernel which uses transferred knowledge from unlabeled data to cope with situations where training examples are scarce. In our approach, unlabeled data has been used to construct an optimized kernel that better generalizes on the target dataset. For the proposed kernel learning algorithm, Fisher Discriminant Analysis (FDA) is used in conjunction with Maximum Mean Discrepancy (MMD) test of statistics to optimize a base kernel using labeled and unlabeled data. Thereafter, the constructed kernel from both labeled and unlabeled datasets is used in SVM to evaluate the results which proved to increase prediction accuracy.

KW - Kernel methods

KW - Learning the kernels

KW - Machine learning

KW - Support Vector Machine (SVM)

KW - Transfer learning

UR - https://www.scopus.com/pages/publications/77949865866

U2 - 10.1109/ICMV.2009.10

DO - 10.1109/ICMV.2009.10

M3 - Conference Paper

AN - SCOPUS:77949865866

SN - 9780769539447

SP - 111

EP - 117

BT - 2009 2nd International Conference on Machine Vision, ICMV 2009

T2 - International Conference on Machine Vision 2009

Y2 - 28 December 2009 through 30 December 2009

ER -

Optimizing kernel functions using transfer learning from unlabeled data

Abstract

Conference

Keywords

Access to Document

Other files and links

Cite this