Skip to content

Initializes sklearn object from flow with default hyperparam configuration - #300

Merged
mfeurer merged 7 commits into
developfrom
init_default_flows
Jun 19, 2018
Merged

Initializes sklearn object from flow with default hyperparam configuration#300
mfeurer merged 7 commits into
developfrom
init_default_flows

Conversation

@janvanrijn

Copy link
Copy Markdown
Member

@mfeurer I am really wondering what you think of the implementation. There is no hurry with this feature

@mfeurer

Copy link
Copy Markdown
Collaborator

Hey, I'd be interested in the rational behind this before actually looking at the implementation.

@janvanrijn

Copy link
Copy Markdown
MemberAuthor

Sklearn has for most hyperparameters some pretty good defaults, which can be a topic of research (e.g., answering the question, how much better are we than the default hyperparameter configuration). This function allows to reinitialize an OpenML flow with these defaults.

The difficulty lies in hyperparameters that do not have any defaults (e.g., the steps of a pipeline). These need to be inferred from the flowname.

@mfeurer

Copy link
Copy Markdown
Collaborator

Okay, so instead of creating the flow via a pipeline with the appropriate hyperparameter configuration, you want to download an existing flow and alter its hyperparameter configuration?

@mfeurer

Copy link
Copy Markdown
Collaborator

Really interesting. If you resolve the merge conflict and update the documentation, I think we can merge this.

@janvanrijn

Copy link
Copy Markdown
MemberAuthor

Thanks for checking this.

I predict that it will work after my most recent commit.

update the documentation

The undocumented part of the function is the 'components' part in kwargs. TBH I have no clue what this does again, and what the rationale is (this function was undocumented before). Can you add it?

@janvanrijn
janvanrijn requested a review from mfeurerJune 18, 2018 18:40
@codecov-io

codecov-io commented Jun 18, 2018

Copy link
Copy Markdown

Codecov Report

Merging #300 into develop will increase coverage by 0.02%.
The diff coverage is 88.88%.

Impacted file tree graph

@@ Coverage Diff @@## develop #300 +/- ##
===========================================
+ Coverage 89.46% 89.48% +0.02% 
===========================================
Files 32 32 Lines 2771 2786 +15 ===========================================
+ Hits 2479 2493 +14 - Misses 292 293 +1
Impacted FilesCoverage Δ
openml/flows/sklearn_converter.py92.2% <88.88%> (+0.04%)⬆️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4603345...bb99ec3. Read the comment docs.

@joaquinvanschoren
joaquinvanschoren self-requested a review June 18, 2018 22:23
@mfeurer
mfeurer merged commit 666d4c7 into developJun 19, 2018
@mfeurer
mfeurer deleted the init_default_flows branch June 19, 2018 08:45
Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants

@janvanrijn@mfeurer@codecov-io@joaquinvanschoren