Scalability: incorporate early pruning optimizations #368
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
* added flag for turning on/off lazy maintain optimization
…xperiment results; Add warning message and test for early pruning
* fixed sampling config test * improved Executor documentation
Codecov Report
@@ Coverage Diff @@
## master #368 +/- ##
==========================================
+ Coverage 84.46% 84.67% +0.20%
==========================================
Files 51 51
Lines 3902 3948 +46
==========================================
+ Hits 3296 3343 +47
+ Misses 606 605 -1
Continue to review full report at Codecov.
|
thyneb19
approved these changes
Apr 28, 2021
There was a problem hiding this comment.
Hey Doris the changes overall look good, also thanks for doing the quick fix for the SQLExecutor. I think there is more work that needs to be done to make the pruning actually compatible with the SQLExecutor case, but maybe we can merge in the PandasExecutor implementation first.
|
Thanks @thyneb19! There definitely needs to be more work to support pruning in SQLExecutor in the future, which would require adding an implementation for |
dorisjlee
added a commit
that referenced
this pull request
Apr 30, 2021
…ales (#262) * Add support to improve temporal action to display different timescales * Resolve PR comments * Add support to improve temporal action to display different timescales * Resolve PR comments * Reformat files using black * "All-column" vis when only few columns in dataframe #199 (#336) Co-authored-by: Caitlyn Chen <caitlynachen@berkeley.edu> Co-authored-by: Doris Lee <dorisjunglinlee@gmail.com> * documentation and cleaning * added notebook gallery * update README * removed scatterplot message in SQLExecutor * fixed typo in SQL documentation * update README and bump version * bump version * clear propagated vis data intent after PandasExecutor completes execute (#297) * fix black to stable version * Scalability: incorporate early pruning optimizations (#368) * changes from perf branch to config * added flag for turning on/off lazy maintain optimization * merged in approx early pruning code * increase overall sampling start and cap * Adjust width and length criteria for early pruning vislist based on experiment results; Add warning message and test for early pruning * black version update * version lock on black * * fixed sql tests (added approx to execute constructor) * fixed sampling config test * improved Executor documentation * timescale feature * adding weekday * adding docs * bugfix for y axis line chart export * fixing temporal axis by adding timescale variable in Clause Co-authored-by: Doris Lee <dorisjunglinlee@gmail.com> Co-authored-by: Caitlyn Chen <caitlynachen@gmail.com> Co-authored-by: Caitlyn Chen <caitlynachen@berkeley.edu>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Overview
Incorporating early pruning optimizations described in our recent paper to speed up cases where the visualization search space is large (e.g., VisList contains more than top K visualizations and the dataframe has a large number of rows).
Changes
execute_approx_samplefor approximating samples for early pruningOthers:
df.unique()Example Output
Here are some results from preliminary experiments on the 1M Airbnb dataset. We duplicate a single visualization N number of times and measuring the performance. The time is the sum of the time it takes to go through the search space and the time that it takes to recompute the top K visualization upon display of the VisList. The significant speedup in the former time supersedes extra time it takes for recompute, around 17, typically no more than a few more than 15 (which k=15).
Zooming out, the speedup can be significant when searching through over 100+ vis:
