ExplainExplore: Visual Exploration of Machine Learning Explanations

Dennis Collaris; Jack J. van Wijk

doi:10.1109/PacificVis48177.2020.7090

ExplainExplore: Visual Exploration of Machine Learning Explanations

Dennis Collaris, Jack J. van Wijk

Visualization

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

45 Citations (Scopus)

1672 Downloads (Pure)

Abstract

Machine learning models often exhibit complex behavior that is difficult to understand. Recent research in explainable AI has produced promising techniques to explain the inner workings of such models using feature contribution vectors. These vectors are helpful in a wide variety of applications. However, there are many parameters involved in this process and determining which settings are best is difficult due to the subjective nature of evaluating interpretability.
To this end, we introduce ExplainExplore: an interactive explanation system to explore explanations that fit the subjective preference of data scientists. We leverage the domain knowledge of the data scientist to find optimal parameter settings and instance perturbations, and enable the discussion of the model and its explanation with domain experts.
We present a use case on a real-world dataset to demonstrate the effectiveness of our approach for the exploration and tuning of machine learning explanations.

Original language	English
Title of host publication	2020 IEEE Pacific Visualization Symposium, PacificVis 2020 - Proceedings
Editors	Fabian Beck, Jinwook Seo, Chaoli Wang
Publisher	Institute of Electrical and Electronics Engineers
Pages	26-35
Number of pages	10
ISBN (Electronic)	978-1-7281-5697-2
DOIs	https://doi.org/10.1109/PacificVis48177.2020.7090
Publication status	Published - 3 Jun 2020
Event	13th IEEE Pacific Visualization Symposium, PacificVis 2020 - Tianjin, China Duration: 14 Apr 2020 → 17 Apr 2020

Conference

Conference	13th IEEE Pacific Visualization Symposium, PacificVis 2020
Country/Territory	China
City	Tianjin
Period	14/04/20 → 17/04/20

Keywords

Computing methodologies
Human-centered computing
Machine learning
Visualization

Access to Document

10.1109/PacificVis48177.2020.7090

paper-pacificvis2020Final author version , 4.04 MB

Cite this

@inproceedings{adfdbcaa50544d6d8cf57239c8aa8c2f,

title = "ExplainExplore: Visual Exploration of Machine Learning Explanations",

abstract = "Machine learning models often exhibit complex behavior that is difficult to understand. Recent research in explainable AI has produced promising techniques to explain the inner workings of such models using feature contribution vectors. These vectors are helpful in a wide variety of applications. However, there are many parameters involved in this process and determining which settings are best is difficult due to the subjective nature of evaluating interpretability.To this end, we introduce ExplainExplore: an interactive explanation system to explore explanations that fit the subjective preference of data scientists. We leverage the domain knowledge of the data scientist to find optimal parameter settings and instance perturbations, and enable the discussion of the model and its explanation with domain experts.We present a use case on a real-world dataset to demonstrate the effectiveness of our approach for the exploration and tuning of machine learning explanations.",

keywords = "Computing methodologies, Human-centered computing, Machine learning, Visualization",

author = "Dennis Collaris and {van Wijk}, {Jack J.}",

year = "2020",

month = jun,

day = "3",

doi = "10.1109/PacificVis48177.2020.7090",

language = "English",

pages = "26--35",

editor = "Fabian Beck and Jinwook Seo and Chaoli Wang",

booktitle = "2020 IEEE Pacific Visualization Symposium, PacificVis 2020 - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

note = "13th IEEE Pacific Visualization Symposium, PacificVis 2020 ; Conference date: 14-04-2020 Through 17-04-2020",

}

Collaris, D & van Wijk, JJ 2020, ExplainExplore: Visual Exploration of Machine Learning Explanations. in F Beck, J Seo & C Wang (eds), 2020 IEEE Pacific Visualization Symposium, PacificVis 2020 - Proceedings., 9086281, Institute of Electrical and Electronics Engineers, pp. 26-35, 13th IEEE Pacific Visualization Symposium, PacificVis 2020, Tianjin, China, 14/04/20. https://doi.org/10.1109/PacificVis48177.2020.7090

ExplainExplore: Visual Exploration of Machine Learning Explanations. / Collaris, Dennis ; van Wijk, Jack J.
2020 IEEE Pacific Visualization Symposium, PacificVis 2020 - Proceedings. ed. / Fabian Beck; Jinwook Seo; Chaoli Wang. Institute of Electrical and Electronics Engineers, 2020. p. 26-35 9086281.

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

TY - GEN

T1 - ExplainExplore: Visual Exploration of Machine Learning Explanations

AU - Collaris, Dennis

AU - van Wijk, Jack J.

PY - 2020/6/3

Y1 - 2020/6/3

N2 - Machine learning models often exhibit complex behavior that is difficult to understand. Recent research in explainable AI has produced promising techniques to explain the inner workings of such models using feature contribution vectors. These vectors are helpful in a wide variety of applications. However, there are many parameters involved in this process and determining which settings are best is difficult due to the subjective nature of evaluating interpretability.To this end, we introduce ExplainExplore: an interactive explanation system to explore explanations that fit the subjective preference of data scientists. We leverage the domain knowledge of the data scientist to find optimal parameter settings and instance perturbations, and enable the discussion of the model and its explanation with domain experts.We present a use case on a real-world dataset to demonstrate the effectiveness of our approach for the exploration and tuning of machine learning explanations.

AB - Machine learning models often exhibit complex behavior that is difficult to understand. Recent research in explainable AI has produced promising techniques to explain the inner workings of such models using feature contribution vectors. These vectors are helpful in a wide variety of applications. However, there are many parameters involved in this process and determining which settings are best is difficult due to the subjective nature of evaluating interpretability.To this end, we introduce ExplainExplore: an interactive explanation system to explore explanations that fit the subjective preference of data scientists. We leverage the domain knowledge of the data scientist to find optimal parameter settings and instance perturbations, and enable the discussion of the model and its explanation with domain experts.We present a use case on a real-world dataset to demonstrate the effectiveness of our approach for the exploration and tuning of machine learning explanations.

KW - Computing methodologies

KW - Human-centered computing

KW - Machine learning

KW - Visualization

UR - http://www.scopus.com/inward/record.url?scp=85085159797&partnerID=8YFLogxK

UR - https://explaining.ml/explainexplore

U2 - 10.1109/PacificVis48177.2020.7090

DO - 10.1109/PacificVis48177.2020.7090

M3 - Conference contribution

SP - 26

EP - 35

BT - 2020 IEEE Pacific Visualization Symposium, PacificVis 2020 - Proceedings

A2 - Beck, Fabian

A2 - Seo, Jinwook

A2 - Wang, Chaoli

PB - Institute of Electrical and Electronics Engineers

T2 - 13th IEEE Pacific Visualization Symposium, PacificVis 2020

Y2 - 14 April 2020 through 17 April 2020

ER -

ExplainExplore: Visual Exploration of Machine Learning Explanations

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

ExplainExplore: Visual Exploration of Machine Learning Explanations (poster)

Cite this