8000 Labels don't stay clamped in LabelPropagation · Issue #3550 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]
Skip to content

Labels don't stay clamped in LabelPropagation #3550

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
mattphotonman opened this issue Aug 11, 2014 · 5 comments
Closed

Labels don't stay clamped in LabelPropagation #3550

mattphotonman opened this issue Aug 11, 2014 · 5 comments
Labels
Bug
Milestone
0.19

Comments

@mattphotonman
Copy link

In some cases the labels don't stay clamped in LabelPropagation. For example:

from sklearn.semi_supervised import LabelPropagation
import numpy as np
lp = LabelPropagation(kernel = 'knn', n_neighbors = 2)
X = np.array([[1.,1.],[1.,0.],[0.,1.]])
y = np.array([1.,0.,-1.])
lp.fit(X,y)
print lp
print y
print lp.transduction_

# Produces:

LabelPropagation(alpha=1, gamma=20, kernel='knn', max_iter=30, n_neighbors=2,
         tol=0.001)
[ 1.  0. -1.]
[ 0.  0.  1.]

I think the problem occurs in label_propagation.py at line 235 and line 249. When alpha is equal to 1, y_static is set to 0 in line 235, and then line 249 doesn't change self.label_distributions_, whereas it should clamp the values of the labelled data points.

My understanding from the documentation is that for LabelPropagation it is always supposed to do hard clamping, and thus it should be completely independent of alpha. I'm not actually sure what the intention was here since the same fit method is used for LabelSpreading.
kpysniak added a commit to kpysniak/scikit-learn that referenced this issue Oct 10, 2014
@kpysniak
[FIX] Fixing Issue scikit-learn#3550 - hard clamping.
7c62fcb
@kpysniak kpysniak mentioned this issue Oct 10, 2014
Closed
@kpysniak
Copy link
Contributor

@mattphotonman I think you're right. It seems like the alpha was not handled correctly and, currently, hard clamping takes place for alpha = 0.

@kpysniak kpysniak mentioned this issue Oct 11, 2014
Closed
@kpysniak
Copy link
Contributor

ping @GaelVaroquaux

@mattphotonman
Copy link
Author

Cool! (Do I have to close this?)

@kpysniak
Copy link
Contributor

If it's verified and confirmed to work, then it should be qualified to be get closed. However, we might wait for @GaelVaroquaux , because he marked the issue as a bug.

@amueller
Copy link
Member

We'll close the issue once the PR is merged. Sorry for the very slow reply.

@amueller amueller added this to the 0.16 milestone Jan 22, 2015
@amueller amueller modified the milestones: 0.16, 0.17 Sep 11, 2015
@amueller amueller modified the milestones: 0.18, 0.17 Sep 20, 2015
@amueller amueller mentioned this issue Nov 20, 2015
Closed
@boechat107 boechat107 mentioned this issue Apr 27, 2016
Closed
@amueller amueller modified the milestones: 0.18, 0.19 Sep 22, 2016
@amueller amueller mentioned this issue Dec 3, 2016
Closed
@jnothman jnothman mentioned this issue Jun 21, 2017
Closed
@musically-ut musically-ut mentioned this issue Jun 28, 2017
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
Bug
Projects
None yet
Milestone
0.19
Development

Successfully merging a pull request may close this issue.

[MRG] Fixing label clamping for LabelPropagation boechat107/scikit-learn
4 participants
@GaelVaroquaux @amueller @kpysniak @mattphotonman
0