Efficient kernel-based learning for trees

Fabio Aiolli, Giovanni Martino, Alessandro Sperduti, Alessandro Moschitti

Research output: Chapter in Book/Report/Conference proceedingConference contribution

11 Citations (Scopus)

Abstract

Kernel methods are effective approaches to the modeling of structured objects in learning algorithms. Their major drawback is the typically high computational complexity of kernel functions. This prevents the application of computational demanding algorithms, e.g. Support Vector Machines, on large dataseis. Consequently, on-line learning approaches are required. Moreover, to facilitate the application of kernel methods on structured data, additional efficiency optimization should be carried out. In this paper, we propose Direct Acyclic Graphs to reduce the computational burden and storage requirements by representing common structures and feature vectors. We show the benefit of our approach for the perceptron algorithm using tree and polynomial kernels. The experiments on a quite extensive dataset of about one million of instances show that our model makes the use of kernels for trees practical. From the accuracy point of view, the possibility of using large amount of data has allowed us to reach the state-of-the-art on the automatic detection of Semantic Role Labeling as defined in the Conference on Natural Language Learning shared task.

Original languageEnglish
Title of host publicationProceedings of the 2007 IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007
Pages308-315
Number of pages8
DOIs
Publication statusPublished - 25 Sep 2007
Externally publishedYes
Event1st IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007 - Honolulu, HI, United States
Duration: 1 Apr 20075 Apr 2007

Other

Other1st IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007
CountryUnited States
CityHonolulu, HI
Period1/4/075/4/07

Fingerprint

Kernel Methods
kernel
Trees (mathematics)
Computational Algorithm
Tree Algorithms
Polynomial Algorithm
Kernel Function
Perceptron
Feature Vector
Natural Language
Labeling
Learning algorithms
Support vector machines
Learning Algorithm
Computational complexity
Support Vector Machine
Computational Complexity
Semantics
Polynomials
Neural networks

ASJC Scopus subject areas

  • Information Systems
  • Signal Processing
  • Software
  • Theoretical Computer Science

Cite this

Aiolli, F., Martino, G., Sperduti, A., & Moschitti, A. (2007). Efficient kernel-based learning for trees. In Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007 (pp. 308-315). [4221313] https://doi.org/10.1109/CIDM.2007.368889

Efficient kernel-based learning for trees. / Aiolli, Fabio; Martino, Giovanni; Sperduti, Alessandro; Moschitti, Alessandro.

Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007. 2007. p. 308-315 4221313.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Aiolli, F, Martino, G, Sperduti, A & Moschitti, A 2007, Efficient kernel-based learning for trees. in Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007., 4221313, pp. 308-315, 1st IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007, Honolulu, HI, United States, 1/4/07. https://doi.org/10.1109/CIDM.2007.368889
Aiolli F, Martino G, Sperduti A, Moschitti A. Efficient kernel-based learning for trees. In Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007. 2007. p. 308-315. 4221313 https://doi.org/10.1109/CIDM.2007.368889
Aiolli, Fabio ; Martino, Giovanni ; Sperduti, Alessandro ; Moschitti, Alessandro. / Efficient kernel-based learning for trees. Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007. 2007. pp. 308-315
@inproceedings{641d64daadba4be0b37c7a5de1570708,
title = "Efficient kernel-based learning for trees",
abstract = "Kernel methods are effective approaches to the modeling of structured objects in learning algorithms. Their major drawback is the typically high computational complexity of kernel functions. This prevents the application of computational demanding algorithms, e.g. Support Vector Machines, on large dataseis. Consequently, on-line learning approaches are required. Moreover, to facilitate the application of kernel methods on structured data, additional efficiency optimization should be carried out. In this paper, we propose Direct Acyclic Graphs to reduce the computational burden and storage requirements by representing common structures and feature vectors. We show the benefit of our approach for the perceptron algorithm using tree and polynomial kernels. The experiments on a quite extensive dataset of about one million of instances show that our model makes the use of kernels for trees practical. From the accuracy point of view, the possibility of using large amount of data has allowed us to reach the state-of-the-art on the automatic detection of Semantic Role Labeling as defined in the Conference on Natural Language Learning shared task.",
author = "Fabio Aiolli and Giovanni Martino and Alessandro Sperduti and Alessandro Moschitti",
year = "2007",
month = "9",
day = "25",
doi = "10.1109/CIDM.2007.368889",
language = "English",
isbn = "1424407052",
pages = "308--315",
booktitle = "Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007",

}

TY - GEN

T1 - Efficient kernel-based learning for trees

AU - Aiolli, Fabio

AU - Martino, Giovanni

AU - Sperduti, Alessandro

AU - Moschitti, Alessandro

PY - 2007/9/25

Y1 - 2007/9/25

N2 - Kernel methods are effective approaches to the modeling of structured objects in learning algorithms. Their major drawback is the typically high computational complexity of kernel functions. This prevents the application of computational demanding algorithms, e.g. Support Vector Machines, on large dataseis. Consequently, on-line learning approaches are required. Moreover, to facilitate the application of kernel methods on structured data, additional efficiency optimization should be carried out. In this paper, we propose Direct Acyclic Graphs to reduce the computational burden and storage requirements by representing common structures and feature vectors. We show the benefit of our approach for the perceptron algorithm using tree and polynomial kernels. The experiments on a quite extensive dataset of about one million of instances show that our model makes the use of kernels for trees practical. From the accuracy point of view, the possibility of using large amount of data has allowed us to reach the state-of-the-art on the automatic detection of Semantic Role Labeling as defined in the Conference on Natural Language Learning shared task.

AB - Kernel methods are effective approaches to the modeling of structured objects in learning algorithms. Their major drawback is the typically high computational complexity of kernel functions. This prevents the application of computational demanding algorithms, e.g. Support Vector Machines, on large dataseis. Consequently, on-line learning approaches are required. Moreover, to facilitate the application of kernel methods on structured data, additional efficiency optimization should be carried out. In this paper, we propose Direct Acyclic Graphs to reduce the computational burden and storage requirements by representing common structures and feature vectors. We show the benefit of our approach for the perceptron algorithm using tree and polynomial kernels. The experiments on a quite extensive dataset of about one million of instances show that our model makes the use of kernels for trees practical. From the accuracy point of view, the possibility of using large amount of data has allowed us to reach the state-of-the-art on the automatic detection of Semantic Role Labeling as defined in the Conference on Natural Language Learning shared task.

UR - http://www.scopus.com/inward/record.url?scp=34548785173&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34548785173&partnerID=8YFLogxK

U2 - 10.1109/CIDM.2007.368889

DO - 10.1109/CIDM.2007.368889

M3 - Conference contribution

SN - 1424407052

SN - 9781424407057

SP - 308

EP - 315

BT - Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Data Mining, CIDM 2007

ER -