Interactive Symbolic Regression - A Study on Noise Sensitivity and Extrapolation Accuracy
Created by W.Langdon from
gp-bibliography.bib Revision:1.7975
- @InProceedings{raghav:2024:GECCOcomp,
-
author = "S. Sanjith Raghav and S. Tejesh Kumar and
Rishiikesh Balaji and M. Sanjay and C. Shunmuga Velayutham",
-
title = "Interactive Symbolic Regression - A Study on Noise
Sensitivity and Extrapolation Accuracy",
-
booktitle = "Symbolic Regression",
-
year = "2024",
-
editor = "William {La Cava} and Steven Gustafson",
-
pages = "2076--2082",
-
address = "Melbourne, Australia",
-
series = "GECCO '24",
-
month = "14-18 " # jul,
-
organisation = "SIGEVO",
-
publisher = "Association for Computing Machinery",
-
publisher_address = "New York, NY, USA",
-
keywords = "genetic algorithms, genetic programming, interactive
symbolic regression, gplearn, user-interactive search
and interactive evolutionary computation",
-
isbn13 = "979-8-4007-0495-6",
-
DOI = "doi:10.1145/3638530.3664130",
-
size = "7 pages",
-
abstract = "This paper presents an interactive symbolic regression
framework i-gplearn, which extends the popular Python
Symbolic Regression library gplearn with user
interactivity. Predominatly, all the Symbolic
Regression (SR) algorithms focus on best model fit
despite the fact that the final objective is more about
scientific insight for the user. Consequently, the
possibility of relying on human intuition and expertise
for exploratory discovery of models has greater
potential in the discovery of relevant and useful
mathematical expressions. i-gplearn combines both user
score and gplearn's model accuracy. Recent benchmark
studies have demonstrated that the state-of-the-art SR
approaches find difficulty in solving problems
sensitive to noise and problems demanding extrapolation
accuracy. i-gplearn interactive experiments with 11
users have been demonstrated on the two tasks -
extrapolation accuracy and sensitive to noise conducted
as part of the Symbolic Regression competition hosted
in 2022 Genetic and Evolutionary Computation Conference
(GECCO). The user studies showed competitive
performance of i-gplearn against the baseline gplearn
in the sensitive to noise task. In the case of
extrapolation accuracy, the user interaction studies
showed both huge potential in identifying diverse
expressions as well as the challenges of interactive
experiments.",
-
notes = "GECCO-2024 SymReg A Recombination of the 33rd
International Conference on Genetic Algorithms (ICGA)
and the 29th Annual Genetic Programming Conference
(GP)",
- }
Genetic Programming entries for
S Sanjith Raghav
S Tejesh Kumar
Rishiikesh Balaji
M Sanjay
C Shunmuga Velayutham
Citations