L1 LASSO Modeling and Its Bayesian Inference

Junbin Gao; Michael Antolovich; Paul Kwan

doi:10.1007/978-3-540-89378-3

L1 LASSO Modeling and Its Bayesian Inference

Junbin Gao, Michael Antolovich, Paul Kwan

Research output: Book chapter/Published conference paper › Conference paper › peer-review

41 Downloads (Pure)

Abstract

A new iterative procedure for solving regression problems with the so-called LASSO penalty [1] is proposed by using generative Bayesian modeling and inference. The algorithm produces the anticipated parsimonious or sparse regression models that generalize well on unseen data. The proposed algorithm is quite robust and there is noneed to specify any model hyperparameters. A comparison with state-of-the-art methods for constructing sparse regression models such as the relevance vector machine (RVM) and the local regularization assisted orthogonal least squares regression (LROLS) is given.

Original language	English
Title of host publication	AI 2008 21st conference
Subtitle of host publication	Advances in artificial intelligence
Editors	W. Wobcke, M. Zhang
Place of Publication	Netherland
Publisher	Springer
Pages	318-324
Number of pages	7
Volume	5360/2008
DOIs	https://doi.org/10.1007/978-3-540-89378-3
Publication status	Published - 2008
Event	Australian Joint Conference on Artificial Intelligence - Auckland, New Zealand, New Zealand Duration: 01 Dec 2008 → 05 Dec 2008

Conference

Conference	Australian Joint Conference on Artificial Intelligence
Country/Territory	New Zealand
Period	01/12/08 → 05/12/08

Access to Document

10.1007/978-3-540-89378-3

CSU358003Accepted author manuscript, 261 KB

Cite this

@inproceedings{1b5423d1c3fc47a2878a91cbac676b43,

title = "L1 LASSO Modeling and Its Bayesian Inference",

abstract = "A new iterative procedure for solving regression problems with the so-called LASSO penalty [1] is proposed by using generative Bayesian modeling and inference. The algorithm produces the anticipated parsimonious or sparse regression models that generalize well on unseen data. The proposed algorithm is quite robust and there is noneed to specify any model hyperparameters. A comparison with state-of-the-art methods for constructing sparse regression models such as the relevance vector machine (RVM) and the local regularization assisted orthogonal least squares regression (LROLS) is given.",

keywords = "Open access version available, Kernel Model, LROLS, Regression",

author = "Junbin Gao and Michael Antolovich and Paul Kwan",

note = "Imported on 03 May 2017 - DigiTool details were: publisher = Netherland: Springer, 2008. editor/s (773b) = W Wobcke and M Zhang; Event dates (773o) = 1-5 Dec 2008; Parent title (773t) = Australian Joint Conference on Artificial Intelligence. ISSNs: 0302-9743; ; Australian Joint Conference on Artificial Intelligence ; Conference date: 01-12-2008 Through 05-12-2008",

year = "2008",

doi = "10.1007/978-3-540-89378-3",

language = "English",

volume = "5360/2008",

pages = "318--324",

editor = "W. Wobcke and M. Zhang",

booktitle = "AI 2008 21st conference",

publisher = "Springer",

address = "United States",

}

TY - GEN

T1 - L1 LASSO Modeling and Its Bayesian Inference

AU - Gao, Junbin

AU - Antolovich, Michael

AU - Kwan, Paul

N1 - Imported on 03 May 2017 - DigiTool details were: publisher = Netherland: Springer, 2008. editor/s (773b) = W Wobcke and M Zhang; Event dates (773o) = 1-5 Dec 2008; Parent title (773t) = Australian Joint Conference on Artificial Intelligence. ISSNs: 0302-9743;

PY - 2008

Y1 - 2008

N2 - A new iterative procedure for solving regression problems with the so-called LASSO penalty [1] is proposed by using generative Bayesian modeling and inference. The algorithm produces the anticipated parsimonious or sparse regression models that generalize well on unseen data. The proposed algorithm is quite robust and there is noneed to specify any model hyperparameters. A comparison with state-of-the-art methods for constructing sparse regression models such as the relevance vector machine (RVM) and the local regularization assisted orthogonal least squares regression (LROLS) is given.

AB - A new iterative procedure for solving regression problems with the so-called LASSO penalty [1] is proposed by using generative Bayesian modeling and inference. The algorithm produces the anticipated parsimonious or sparse regression models that generalize well on unseen data. The proposed algorithm is quite robust and there is noneed to specify any model hyperparameters. A comparison with state-of-the-art methods for constructing sparse regression models such as the relevance vector machine (RVM) and the local regularization assisted orthogonal least squares regression (LROLS) is given.

KW - Open access version available

KW - Kernel Model

KW - LROLS

KW - Regression

U2 - 10.1007/978-3-540-89378-3

DO - 10.1007/978-3-540-89378-3

M3 - Conference paper

VL - 5360/2008

SP - 318

EP - 324

BT - AI 2008 21st conference

A2 - Wobcke, W.

A2 - Zhang, M.

PB - Springer

CY - Netherland

T2 - Australian Joint Conference on Artificial Intelligence

Y2 - 1 December 2008 through 5 December 2008

ER -

L1 LASSO Modeling and Its Bayesian Inference

Abstract

Conference

Access to Document

Fingerprint

Cite this