Thanks to visit codestin.com
Credit goes to github.com

Skip to content

log-responsibilities in GMM #3813

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
superbobry opened this issue Oct 30, 2014 · 5 comments
Closed

log-responsibilities in GMM #3813

superbobry opened this issue Oct 30, 2014 · 5 comments
Labels
Enhancement

Comments

@superbobry
Copy link
Contributor

Hello. The current implementation of GMM.score_samples computes log-responsibilities, exponentiates them and then returns only the exponentiated version. I wonder if there is there a reason for that?

Having a way to extract log-responsibilities might improve the accuracy of more complex models built on top of GMM, for example GMMHMM.
@agramfort
Copy link
Member

the GMM module needs some love. Any help welcome.

@superbobry
Copy link
Contributor Author

I've looked through the models in naive_bayes since they are also generative. BaseNB doesn't define score_samples, but instead it has two methods for obtaining posterior probabilities: predict_proba and predict_log_proba.

I can submit a PR unifying the API of GMM and DPGMM with naive_bayes models.

@agramfort
Copy link
Member

fine but keep in mind that GMM are unsupervised models not like naive_bayes

@amueller
Copy link
Member

Also see #4062

@tguillemot
Copy link
Contributor

Solved by #6666.
Now score sample returns log-responsabilities and have a predict_proba method

@amueller amueller closed this as completed Oct 5, 2016
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
Enhancement
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
4 participants
@agramfort @superbobry @amueller @tguillemot