Evaluating layer-wise relevance propagation explainability maps for artificial neural networks

Elena Ranguelova; Eric J. Pauwels; Joost Berkhout

doi:10.1109/eScience.2018.00107

Evaluating layer-wise relevance propagation explainability maps for artificial neural networks

Elena Ranguelova, Eric J. Pauwels, Joost Berkhout

Mathematics

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

57 Downloads (Pure)

Abstract

Layer-wise relevance propagation (LRP) heatmaps aim to provide graphical explanation for decisions of a classifier. This could be of great benefit to scientists for trusting complex black-box models and getting insights from their data. The LRP heatmaps tested on benchmark datasets are reported to correlate significantly with interpretable image features. In this work, we investigate these claims and propose to refine them.

Original language	English
Title of host publication	2018 IEEE 14th International Conference on eScience (e-Science)
Subtitle of host publication	[Proceedings]
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	377-378
Number of pages	2
ISBN (Electronic)	9781538691564
ISBN (Print)	9781538691571
DOIs	https://doi.org/10.1109/eScience.2018.00107
Publication status	Published - 2018
Event	14th IEEE International Conference on eScience, e-Science 2018 - Amsterdam, Netherlands Duration: 29 Oct 2018 → 1 Nov 2018

Conference

Conference	14th IEEE International Conference on eScience, e-Science 2018
Country/Territory	Netherlands
City	Amsterdam
Period	29/10/18 → 1/11/18

UN SDGs

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

Access to Document

10.1109/eScience.2018.00107

Evaluating layer-wise relevance propagation explainability maps for artificial neural networksFinal published version, 169 KBLicence: Article 25fa Dutch Copyright Act

Persistent URL (handle)

Persistent link

Cite this

@inproceedings{c70f1c0b0f744d08a1e0f87cf0581011,

title = "Evaluating layer-wise relevance propagation explainability maps for artificial neural networks",

abstract = "Layer-wise relevance propagation (LRP) heatmaps aim to provide graphical explanation for decisions of a classifier. This could be of great benefit to scientists for trusting complex black-box models and getting insights from their data. The LRP heatmaps tested on benchmark datasets are reported to correlate significantly with interpretable image features. In this work, we investigate these claims and propose to refine them.",

author = "Elena Ranguelova and Pauwels, {Eric J.} and Joost Berkhout",

year = "2018",

doi = "10.1109/eScience.2018.00107",

language = "English",

isbn = "9781538691571",

pages = "377--378",

booktitle = "2018 IEEE 14th International Conference on eScience (e-Science)",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

note = "14th IEEE International Conference on eScience, e-Science 2018 ; Conference date: 29-10-2018 Through 01-11-2018",

}

Ranguelova, E, Pauwels, EJ & Berkhout, J 2018, Evaluating layer-wise relevance propagation explainability maps for artificial neural networks. in 2018 IEEE 14th International Conference on eScience (e-Science): [Proceedings]., 8588726, Institute of Electrical and Electronics Engineers Inc., pp. 377-378, 14th IEEE International Conference on eScience, e-Science 2018, Amsterdam, Netherlands, 29/10/18. https://doi.org/10.1109/eScience.2018.00107

Evaluating layer-wise relevance propagation explainability maps for artificial neural networks. / Ranguelova, Elena; Pauwels, Eric J.; Berkhout, Joost.
2018 IEEE 14th International Conference on eScience (e-Science): [Proceedings]. Institute of Electrical and Electronics Engineers Inc., 2018. p. 377-378 8588726.

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

TY - GEN

T1 - Evaluating layer-wise relevance propagation explainability maps for artificial neural networks

AU - Ranguelova, Elena

AU - Pauwels, Eric J.

AU - Berkhout, Joost

PY - 2018

Y1 - 2018

N2 - Layer-wise relevance propagation (LRP) heatmaps aim to provide graphical explanation for decisions of a classifier. This could be of great benefit to scientists for trusting complex black-box models and getting insights from their data. The LRP heatmaps tested on benchmark datasets are reported to correlate significantly with interpretable image features. In this work, we investigate these claims and propose to refine them.

AB - Layer-wise relevance propagation (LRP) heatmaps aim to provide graphical explanation for decisions of a classifier. This could be of great benefit to scientists for trusting complex black-box models and getting insights from their data. The LRP heatmaps tested on benchmark datasets are reported to correlate significantly with interpretable image features. In this work, we investigate these claims and propose to refine them.

UR - http://www.scopus.com/inward/record.url?scp=85061401855&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85061401855&partnerID=8YFLogxK

U2 - 10.1109/eScience.2018.00107

DO - 10.1109/eScience.2018.00107

M3 - Conference contribution

AN - SCOPUS:85061401855

SN - 9781538691571

SP - 377

EP - 378

BT - 2018 IEEE 14th International Conference on eScience (e-Science)

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 14th IEEE International Conference on eScience, e-Science 2018

Y2 - 29 October 2018 through 1 November 2018

ER -

Evaluating layer-wise relevance propagation explainability maps for artificial neural networks

Abstract

Conference

UN SDGs

Access to Document

Persistent URL (handle)

Other files and links

Fingerprint

Cite this