Spatio-Temporal Convolution Kernels

Publikation: Beiträge in ZeitschriftenZeitschriftenaufsätzeForschungbegutachtet

Standard

Spatio-Temporal Convolution Kernels. / Knauf, Konstantin; Memmert, Daniel; Brefeld, Ulf.
in: Machine Learning, Jahrgang 102, Nr. 2, 01.02.2016, S. 247-273.

Publikation: Beiträge in ZeitschriftenZeitschriftenaufsätzeForschungbegutachtet

Harvard

APA

Vancouver

Knauf K, Memmert D, Brefeld U. Spatio-Temporal Convolution Kernels. Machine Learning. 2016 Feb 1;102(2):247-273. doi: 10.1007/s10994-015-5520-1

Bibtex

@article{44a736a130f64aeeb751164c3ee62e3e,
title = "Spatio-Temporal Convolution Kernels",
abstract = "Trajectory data of simultaneously moving objects is being recorded in many different domains and applications. However, existing techniques that utilise such data often fail to capture characteristic traits or lack theoretical guarantees. We propose a novel class of spatio-temporal convolution kernels to capture similarities in multi-object scenarios. The abstract kernel is a composition of a temporal and a spatial kernel and its actual instantiations depend on the application at hand. Empirically, we compare our kernels and efficient approximations thereof to baseline techniques for clustering tasks using artificial and real world data from team sports.",
keywords = "Engineering, Convolution kernel, Spatio-temporal, Trajectory, Soccer, Business informatics",
author = "Konstantin Knauf and Daniel Memmert and Ulf Brefeld",
year = "2016",
month = feb,
day = "1",
doi = "10.1007/s10994-015-5520-1",
language = "English",
volume = "102",
pages = "247--273",
journal = "Machine Learning",
issn = "0885-6125",
publisher = "Springer US",
number = "2",

}

RIS

TY - JOUR

T1 - Spatio-Temporal Convolution Kernels

AU - Knauf, Konstantin

AU - Memmert, Daniel

AU - Brefeld, Ulf

PY - 2016/2/1

Y1 - 2016/2/1

N2 - Trajectory data of simultaneously moving objects is being recorded in many different domains and applications. However, existing techniques that utilise such data often fail to capture characteristic traits or lack theoretical guarantees. We propose a novel class of spatio-temporal convolution kernels to capture similarities in multi-object scenarios. The abstract kernel is a composition of a temporal and a spatial kernel and its actual instantiations depend on the application at hand. Empirically, we compare our kernels and efficient approximations thereof to baseline techniques for clustering tasks using artificial and real world data from team sports.

AB - Trajectory data of simultaneously moving objects is being recorded in many different domains and applications. However, existing techniques that utilise such data often fail to capture characteristic traits or lack theoretical guarantees. We propose a novel class of spatio-temporal convolution kernels to capture similarities in multi-object scenarios. The abstract kernel is a composition of a temporal and a spatial kernel and its actual instantiations depend on the application at hand. Empirically, we compare our kernels and efficient approximations thereof to baseline techniques for clustering tasks using artificial and real world data from team sports.

KW - Engineering

KW - Convolution kernel

KW - Spatio-temporal

KW - Trajectory

KW - Soccer

KW - Business informatics

UR - http://www.scopus.com/inward/record.url?scp=84955642460&partnerID=8YFLogxK

U2 - 10.1007/s10994-015-5520-1

DO - 10.1007/s10994-015-5520-1

M3 - Journal articles

VL - 102

SP - 247

EP - 273

JO - Machine Learning

JF - Machine Learning

SN - 0885-6125

IS - 2

ER -

DOI