FIX Fixes issue with exatly_zero_info_score #19179
Merged
Conversation
This file contains hidden or 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
|
@thomasjpfan How did you reproduce the error. I installed the same library version as on the CI and I could not reproduce it. Would it be linked to a different compiler? |
|
I was never able to reproduce the error locally. I had to slowly use the CI to figure out where the issue was. |
|
Thanks for this :)
…On Thu, 14 Jan 2021 at 22:53, Thomas J. Fan ***@***.***> wrote:
I was never able to reproduce the error locally. I had to slowly use the
CI to figure out where the issue was.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#19179 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABY32P3PHDWWEILFDIDILY3SZ5RXLANCNFSM4WDDNLGQ>
.
--
Guillaume Lemaitre
Scikit-learn @ Inria Foundation
https://glemaitre.github.io/
|
|
Even so, the issue persist. I think I can create a small test example using |
|
It seems that this is only an issue in openml now? |
thomasjpfan
commented
Jan 15, 2021
Comment on lines
+41
to
+42
| log_a = np.log(a) | ||
| log_b = np.log(b) |
There was a problem hiding this comment.
Slightly more memory efficient because we would not need to create the 2d array anymore.
lorentzenchr
approved these changes
Jan 16, 2021
| @@ -38,9 +38,10 @@ def expected_mutual_information(contingency, int n_samples): | |||
| term1 = nijs / N | |||
| # term2 is log((N*nij) / (a * b)) == log(N * nij) - log(a * b) | |||
| # term2 uses the outer product | |||
There was a problem hiding this comment.
Could you adapt/remove this comment. With this PR, no outer product anymore.
adrinjalali
approved these changes
Jan 16, 2021
| @@ -795,6 +795,7 @@ def mutual_info_score(labels_true, labels_pred, *, contingency=None): | |||
| log_outer = -np.log(outer) + log(pi.sum()) + log(pj.sum()) | |||
| mi = (contingency_nm * (log_contingency_nm - log(contingency_sum)) + | |||
| contingency_nm * log_outer) | |||
| mi = np.where(np.abs(mi) < np.finfo(mi.dtype).eps, 0.0, mi) | |||
There was a problem hiding this comment.
fair enough, but I wonder what other places we need to be doing it!
| @@ -54,12 +54,12 @@ def expected_mutual_information(contingency, int n_samples): | |||
| start = np.maximum(start, 1) | |||
| end = np.minimum(np.resize(a, (C, R)).T, np.resize(b, (R, C))) + 1 | |||
| # emi itself is a summation over the various values. | |||
| emi = 0 | |||
| emi = 0.0 | |||
There was a problem hiding this comment.
emi is defined as DOUBLE, not sure why you've added the .0
There was a problem hiding this comment.
it is more explicit if you did not read the definition :)
glemaitre
pushed a commit
to glemaitre/scikit-learn
that referenced
this pull request
Jan 18, 2021
* ENH Fixes issue with exatly_zero_info_score [scipy-dev] * ENH Remove unneeded line [scipy-dev] * WIP Keep types [scipy-dev] * REV Smaller diff [scipy-dev] * WIP Expand mutual_info_score [scipy-dev] * WIP Removes float casting [scipy-dev] * WIP Adds casting back in * CI [scipy-dev] * WIP Casting is not needed [scipy-dev] * WIP Only clip [scipy-dev] * REV Smaller diff [scipy-dev] * WIP Place expected_mutual_information diff back [scipy-dev] * ENH Uses around * WIP Use where again [scipy-dev] * ENH Adjust comments to match code
jeremiedbb
pushed a commit
that referenced
this pull request
Jan 19, 2021
* ENH Fixes issue with exatly_zero_info_score [scipy-dev] * ENH Remove unneeded line [scipy-dev] * WIP Keep types [scipy-dev] * REV Smaller diff [scipy-dev] * WIP Expand mutual_info_score [scipy-dev] * WIP Removes float casting [scipy-dev] * WIP Adds casting back in * CI [scipy-dev] * WIP Casting is not needed [scipy-dev] * WIP Only clip [scipy-dev] * REV Smaller diff [scipy-dev] * WIP Place expected_mutual_information diff back [scipy-dev] * ENH Uses around * WIP Use where again [scipy-dev] * ENH Adjust comments to match code
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.
Reference Issues/PRs
Fixes #19165
What does this implement/fix? Explain your changes.
The issue stemmed from an numerical error where
log_Ndid not cancel outlog_a