Recently, researchers often fail to reproduce published study results for
![]() [Kaiser 2015] |
The Reasons
|
of sequential and linked
“The two important features of the provenance of a data product are theancestral data product(s) from which this data product evolved, and the process oftransformation of these ancestral data product(s), potentially through workflows, that helpedderive this data product. ”
[Simmhan et al. 2005]
Visual analysis of
a web-based data
Refinery Workflow Visualization
refinery-platform.org
|
|
![]() |
![]() |
Workflow
Dataset + multiple analyses =>
Interactive provenance
from the perspective of an Analyst
Node-Link
![]() |
Dynamic Graph Layout
![]() |
Filtering
![]() |
Aggregation ![]() |
Motif Compression ![]() |
Degree-of-Interest ![]() |
Analysis Timeline
![]() |
Glyph Design
![]() |
Path Highlighting
![]() |
![]() |
Workflow Level |
![]() |
Subanalysis Level |
![]() |
Analysis Level |
![]() |
Subanalysis Level |
![]() |
Analysis Level |
![]() |
Layer Level |
Discover analyses with the
Layer analyses with the
Compute deviating amount of incoming or outgoing links and subanalyses.
Subanalyses
Analyses
Layers
Every
General interest: |
|
(v,w) <- [0..1] |
User actions: |
|
v <- [0,1], w <- [0..1] |
Adjust weights: ![]() |
DOI(n) affects node visibility: |
Figure shows the visualization state after step 4
Won
![]() |
![]() |
Analysts are now able to
![]() |
![]() |
![]() |
![]() |
![]() |
Stefan Luger | Holger Stitz | Samuel Gratzl | Nils Gehlenborg | Marc Streit |
Please see thesis bibliography.
Provenance graphs
Filter options: (a) blend, (b) hide.
Graph shows blend and hide for filter option (c).
The
The user wants to investigate the files and transformations leading to a correctly reproduced result.