Modular Dimensionality Reduction

Henry W J Reeve; Tingting Mu; Gavin Brown

doi:10.1007/978-3-030-10925-7_37

Modular Dimensionality Reduction

Henry W J Reeve, Tingting Mu, Gavin Brown

Research output: Chapter in Book/Conference proceeding › Conference contribution › peer-review

141 Downloads (Pure)

Abstract

We introduce an approach to modular dimensionality reduction, allowing efficient learning of multiple complementary representations of the same object. Modules are trained by optimising an unsupervised cost function which balances two competing goals: Maintaining the inner product structure within the original space, and encouraging structural diversity between complementary representations. We derive an efficient learning algorithm which outperforms gradient based approaches without the need to choose a learning rate. We also demonstrate an
intriguing connection with Dropout. Empirical results demonstrate the efficacy of the method for image retrieval and classification.

Original language	English
Title of host publication	Proceedings of the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases 2018
DOIs	https://doi.org/10.1007/978-3-030-10925-7_37
Publication status	Published - 2019

Keywords

Ensemble learning
Dimensionality reduction
Dropout
Kernel principal components analysis

Access to Document

10.1007/978-3-030-10925-7_37

ModularDimensionalityReductionECML18 (7)Accepted author manuscript, 625 KB

Cite this

@inproceedings{dadcca03fd39435c923a4ea85453431a,

title = "Modular Dimensionality Reduction",

abstract = "We introduce an approach to modular dimensionality reduction, allowing efficient learning of multiple complementary representations of the same object. Modules are trained by optimising an unsupervised cost function which balances two competing goals: Maintaining the inner product structure within the original space, and encouraging structural diversity between complementary representations. We derive an efficient learning algorithm which outperforms gradient based approaches without the need to choose a learning rate. We also demonstrate anintriguing connection with Dropout. Empirical results demonstrate the efficacy of the method for image retrieval and classification.",

keywords = "Ensemble learning, Dimensionality reduction, Dropout, Kernel principal components analysis",

author = "Reeve, {Henry W J} and Tingting Mu and Gavin Brown",

year = "2019",

doi = "10.1007/978-3-030-10925-7_37",

language = "English",

booktitle = "Proceedings of the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases 2018",

}

TY - GEN

T1 - Modular Dimensionality Reduction

AU - Reeve, Henry W J

AU - Mu, Tingting

AU - Brown, Gavin

PY - 2019

Y1 - 2019

N2 - We introduce an approach to modular dimensionality reduction, allowing efficient learning of multiple complementary representations of the same object. Modules are trained by optimising an unsupervised cost function which balances two competing goals: Maintaining the inner product structure within the original space, and encouraging structural diversity between complementary representations. We derive an efficient learning algorithm which outperforms gradient based approaches without the need to choose a learning rate. We also demonstrate anintriguing connection with Dropout. Empirical results demonstrate the efficacy of the method for image retrieval and classification.

AB - We introduce an approach to modular dimensionality reduction, allowing efficient learning of multiple complementary representations of the same object. Modules are trained by optimising an unsupervised cost function which balances two competing goals: Maintaining the inner product structure within the original space, and encouraging structural diversity between complementary representations. We derive an efficient learning algorithm which outperforms gradient based approaches without the need to choose a learning rate. We also demonstrate anintriguing connection with Dropout. Empirical results demonstrate the efficacy of the method for image retrieval and classification.

KW - Ensemble learning

KW - Dimensionality reduction

KW - Dropout

KW - Kernel principal components analysis

U2 - 10.1007/978-3-030-10925-7_37

DO - 10.1007/978-3-030-10925-7_37

M3 - Conference contribution

BT - Proceedings of the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases 2018

ER -

Modular Dimensionality Reduction

Abstract

Keywords

Access to Document

Fingerprint

Cite this