Learning Transformation Invariance from Global to Local

Jens Hocke; Thomas Martinetz

Learning Transformation Invariance from Global to Local

Institute of Neuro- and Bioinformatics

Abstract

Learning representations invariant to image transformations is fundamental to improving object recognition. We explore the connections between i-theory, Toroidal Subspace Analysis and slow subspace learning. All these methods can only achieve invariance to one transformation. Motivated by this limitation of these global methods we adapt the slow subspace approach to a local convolutional setting. Experimentally we show invariance to multiple transformations, and test object recognition performance.

Original language	English
Title of host publication	Workshop New Challenges in Neural Computation 2015
Editors	Barbara Hammer, Thomas Villmann
Number of pages	9
Volume	03
Publication date	2015
Pages	16-24
Publication status	Published - 2015
Event	37th German Conference on Pattern Recognition - Workshop Program: New Challenges in Neural Computation (NC2) - RWTH Aachen, Aachen, Germany Duration: 10.10.2015 → 10.10.2015 http://vmv2015.rwth-aachen.de/workshops.html

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being
SDG 9 Industry, Innovation, and Infrastructure

Access to Document

http://www.inb.uni-luebeck.de/publications/pdfs/HoMa15.pdf

Cite this

@inproceedings{ad5a63f405e24c3abed530c81f12429b,

title = "Learning Transformation Invariance from Global to Local",

abstract = "Learning representations invariant to image transformations is fundamental to improving object recognition. We explore the connections between i-theory, Toroidal Subspace Analysis and slow subspace learning. All these methods can only achieve invariance to one transformation. Motivated by this limitation of these global methods we adapt the slow subspace approach to a local convolutional setting. Experimentally we show invariance to multiple transformations, and test object recognition performance.",

author = "Jens Hocke and Thomas Martinetz",

year = "2015",

language = "English",

volume = "03",

series = "Machine Learning Reports",

pages = "16--24",

editor = "Barbara Hammer and Thomas Villmann",

booktitle = "Workshop New Challenges in Neural Computation 2015",

note = "37th German Conference on Pattern Recognition - Workshop Program ; Conference date: 10-10-2015 Through 10-10-2015",

url = "http://vmv2015.rwth-aachen.de/workshops.html",

}

Hocke, J & Martinetz, T 2015, Learning Transformation Invariance from Global to Local. in B Hammer & T Villmann (eds), Workshop New Challenges in Neural Computation 2015. vol. 03, Machine Learning Reports, pp. 16-24, 37th German Conference on Pattern Recognition - Workshop Program, Aachen, Germany, 10.10.15. <http://www.inb.uni-luebeck.de/publications/pdfs/HoMa15.pdf>

TY - GEN

T1 - Learning Transformation Invariance from Global to Local

AU - Hocke, Jens

AU - Martinetz, Thomas

PY - 2015

Y1 - 2015

N2 - Learning representations invariant to image transformations is fundamental to improving object recognition. We explore the connections between i-theory, Toroidal Subspace Analysis and slow subspace learning. All these methods can only achieve invariance to one transformation. Motivated by this limitation of these global methods we adapt the slow subspace approach to a local convolutional setting. Experimentally we show invariance to multiple transformations, and test object recognition performance.

AB - Learning representations invariant to image transformations is fundamental to improving object recognition. We explore the connections between i-theory, Toroidal Subspace Analysis and slow subspace learning. All these methods can only achieve invariance to one transformation. Motivated by this limitation of these global methods we adapt the slow subspace approach to a local convolutional setting. Experimentally we show invariance to multiple transformations, and test object recognition performance.

M3 - Conference contribution

VL - 03

T3 - Machine Learning Reports

SP - 16

EP - 24

BT - Workshop New Challenges in Neural Computation 2015

A2 - Hammer, Barbara

A2 - Villmann, Thomas

T2 - 37th German Conference on Pattern Recognition - Workshop Program

Y2 - 10 October 2015 through 10 October 2015

ER -