Incremental Search Space Construction for Machine Learning Pipeline Synthesis

Zöller, M-A; Nguyen, T-D; Huber, MF

Incremental Search Space Construction for Machine Learning Pipeline Synthesis

Zöller, M-A Nguyen, T-D Huber, MF

Permalink

Publisher:: Springer
Publication Type:: Conference Proceeding
Citation:: Advances in Intelligent Data Analysis XIX, 2021, 12695, pp. 103-115
Issue Date:: 2021-04-13

Closed Access

	Filename	Description	Size
	Zöller2021_Chapter_IncrementalSearchSpaceConstruc.pdf	Published version	488.38 kB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Closed Access

This item is closed access and not available.

Full metadata record

Field	Value	Language
dc.contributor.author	Zöller, M-A
dc.contributor.author	Nguyen, T-D
dc.contributor.author	Huber, MF
dc.date	2021-04-26
dc.date.accessioned	2022-07-05T00:30:57Z
dc.date.available	2022-07-05T00:30:57Z
dc.date.issued	2021-04-13
dc.identifier.citation	Advances in Intelligent Data Analysis XIX, 2021, 12695, pp. 103-115
dc.identifier.isbn	978-3-030-74251-5
dc.identifier.issn	0302-9743
dc.identifier.issn	1611-3349
dc.identifier.uri	http://hdl.handle.net/10453/158636
dc.description.abstract	Automated machine learning (AutoML) aims for constructing machine learning (ML) pipelines automatically. Many studies have investigated efficient methods for algorithm selection and hyperparameter optimization. However, methods for ML pipeline synthesis and optimization considering the impact of complex pipeline structures containing multiple preprocessing and classification algorithms have not been studied thoroughly. In this paper, we propose a data-centric approach based on meta-features for pipeline construction and hyperparameter optimization inspired by human behavior. By expanding the pipeline search space incrementally in combination with meta-features of intermediate data sets, we are able to prune the pipeline structure search space efficiently. Consequently, flexible and data set specific ML pipelines can be constructed. We prove the effectiveness and competitiveness of our approach on 28 data sets used in well-established AutoML benchmarks in comparison with state-of-the-art AutoML frameworks.
dc.language	en
dc.publisher	Springer
dc.relation.ispartof	Advances in Intelligent Data Analysis XIX
dc.relation.ispartof	International Symposium on Intelligent Data Analysis
dc.relation.ispartofseries	Lecture Notes in Computer Science
dc.relation.isbasedon	10.1007/978-3-030-74251-5_9
dc.rights	info:eu-repo/semantics/closedAccess
dc.subject.classification	Artificial Intelligence & Image Processing
dc.title	Incremental Search Space Construction for Machine Learning Pipeline Synthesis
dc.type	Conference Proceeding
utslib.citation.volume	12695
utslib.location.activity	Porto, Portugal
pubs.organisational-group	/University of Technology Sydney
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology/School of Computer Science
utslib.copyright.status	closed_access	*
pubs.consider-herdc	false
dc.date.updated	2022-07-05T00:30:56Z
pubs.finish-date	2021-04-28
pubs.place-of-publication	Switzerland
pubs.publication-status	Published
pubs.start-date	2021-04-26
pubs.volume	12695
dc.location	Switzerland

Abstract:

Automated machine learning (AutoML) aims for constructing machine learning (ML) pipelines automatically. Many studies have investigated efficient methods for algorithm selection and hyperparameter optimization. However, methods for ML pipeline synthesis and optimization considering the impact of complex pipeline structures containing multiple preprocessing and classification algorithms have not been studied thoroughly. In this paper, we propose a data-centric approach based on meta-features for pipeline construction and hyperparameter optimization inspired by human behavior. By expanding the pipeline search space incrementally in combination with meta-features of intermediate data sets, we are able to prune the pipeline structure search space efficiently. Consequently, flexible and data set specific ML pipelines can be constructed. We prove the effectiveness and competitiveness of our approach on 28 data sets used in well-established AutoML benchmarks in comparison with state-of-the-art AutoML frameworks.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/158636