MDL-based fitness for feature construction

Leila Shila Shafti; Eduardo Pérez

doi:10.1145/1276958.1277334

MDL-based fitness for feature construction

Leila Shila Shafti^*, Eduardo Pérez

^*Corresponding author for this work

Computer And Information Sciences

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution book

Abstract

Primitive data representation of real-world data facilitates attribute interactions, which make information opaque to most learners. Feature Construction (FC) aims to abstract and encapsulate interactions into new features and outline them to the learner. When a GA is applied to perform FC, the goal is to generate features that facilitate more accurate learning. Then the GA's fitness function should estimate the quality of the constructed features. We propose a new fitness function based on Minimum Description Length (MDL). This fitness is incorporated in MFE2/GA to improve its accuracy. The new system is compared with other systems based on Entropy or error-rate fitness.

Original language	English
Title of host publication	Proceedings of GECCO 2007
Subtitle of host publication	Genetic and Evolutionary Computation Conference
Publisher	Association for Computing Machinery (ACM)
Pages	1875
Number of pages	1
ISBN (Print)	1595936971, 9781595936974
DOIs	https://doi.org/10.1145/1276958.1277334
Publication status	Published - 7 Jul 2007
Event	9th Annual Genetic and Evolutionary Computation Conference, GECCO 2007 - London, United Kingdom Duration: 7 Jul 2007 → 11 Jul 2007

Conference

Conference	9th Annual Genetic and Evolutionary Computation Conference, GECCO 2007
Country/Territory	United Kingdom
City	London
Period	7/07/07 → 11/07/07

Funding

This work is supported by the Spanish Ministry of Science and Technology, Grant number TSI2005-08225-C07-06.

Keywords

Attribute interaction
Entropy
Feature construction
Fitness function
Machine learning
MDL

Access to Document

10.1145/1276958.1277334

https://www.cs.york.ac.uk/rts/docs/GECCO_2007/docs/p1875.pdf

Cite this

@inproceedings{18ef118c262544de8a55be9dbd1464c0,

title = "MDL-based fitness for feature construction",

abstract = "Primitive data representation of real-world data facilitates attribute interactions, which make information opaque to most learners. Feature Construction (FC) aims to abstract and encapsulate interactions into new features and outline them to the learner. When a GA is applied to perform FC, the goal is to generate features that facilitate more accurate learning. Then the GA's fitness function should estimate the quality of the constructed features. We propose a new fitness function based on Minimum Description Length (MDL). This fitness is incorporated in MFE2/GA to improve its accuracy. The new system is compared with other systems based on Entropy or error-rate fitness.",

keywords = "Attribute interaction, Entropy, Feature construction, Fitness function, Machine learning, MDL",

author = "Shafti, {Leila Shila} and Eduardo P{\'e}rez",

year = "2007",

month = jul,

day = "7",

doi = "10.1145/1276958.1277334",

language = "English",

isbn = "1595936971",

pages = "1875",

booktitle = "Proceedings of GECCO 2007",

publisher = "Association for Computing Machinery (ACM)",

address = "United States",

note = "9th Annual Genetic and Evolutionary Computation Conference, GECCO 2007 ; Conference date: 07-07-2007 Through 11-07-2007",

}

TY - GEN

T1 - MDL-based fitness for feature construction

AU - Shafti, Leila Shila

AU - Pérez, Eduardo

PY - 2007/7/7

Y1 - 2007/7/7

N2 - Primitive data representation of real-world data facilitates attribute interactions, which make information opaque to most learners. Feature Construction (FC) aims to abstract and encapsulate interactions into new features and outline them to the learner. When a GA is applied to perform FC, the goal is to generate features that facilitate more accurate learning. Then the GA's fitness function should estimate the quality of the constructed features. We propose a new fitness function based on Minimum Description Length (MDL). This fitness is incorporated in MFE2/GA to improve its accuracy. The new system is compared with other systems based on Entropy or error-rate fitness.

AB - Primitive data representation of real-world data facilitates attribute interactions, which make information opaque to most learners. Feature Construction (FC) aims to abstract and encapsulate interactions into new features and outline them to the learner. When a GA is applied to perform FC, the goal is to generate features that facilitate more accurate learning. Then the GA's fitness function should estimate the quality of the constructed features. We propose a new fitness function based on Minimum Description Length (MDL). This fitness is incorporated in MFE2/GA to improve its accuracy. The new system is compared with other systems based on Entropy or error-rate fitness.

KW - Attribute interaction

KW - Entropy

KW - Feature construction

KW - Fitness function

KW - Machine learning

KW - MDL

UR - http://www.scopus.com/inward/record.url?scp=34548098108&partnerID=8YFLogxK

U2 - 10.1145/1276958.1277334

DO - 10.1145/1276958.1277334

M3 - Conference contribution book

AN - SCOPUS:34548098108

SN - 1595936971

SN - 9781595936974

SP - 1875

BT - Proceedings of GECCO 2007

PB - Association for Computing Machinery (ACM)

T2 - 9th Annual Genetic and Evolutionary Computation Conference, GECCO 2007

Y2 - 7 July 2007 through 11 July 2007

ER -

MDL-based fitness for feature construction

Abstract

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this