Skip to content
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.

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

add multiclass support for hyperopt utl #1682

Merged
merged 7 commits into from
Jan 11, 2019

Conversation

awant
Copy link
Contributor

@awant awant commented Nov 24, 2018

add support for multiclass classification problems.
Also, find out that we have to speed up script to make it actually useful.

@awant awant changed the title add multiclass support add multiclass support for hyperopt utl Nov 24, 2018
@@ -219,11 +220,16 @@ def get_hyperparam_string(self, **kwargs):
def compose_vw_train_command(self):
data_part = ('vw -d %s -f %s --holdout_off -c '
% (self.train_set, self.train_model))
if self.labels_clf_count > 2: # multiclass, should take probabilities
data_part += ('--oaa %s --loss_function=logistic --probabilities '
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would be nice to support other multiclass methods in addition to --oaa.
e.g.: --ect, --recall_tree, --wap, (with & without ldf) etc.
They all have similar interface (expect number of classes as arg)

@JohnLangford
Copy link
Member

I'm going to merge this because it seems better to have some multiclass support than none. @arielf or others, feel free to tweak to add more multiclass support.

@JohnLangford JohnLangford merged commit b4cafe3 into VowpalWabbit:master Jan 11, 2019
jackdoe pushed a commit to lucjb/vowpal_wabbit that referenced this pull request Mar 7, 2019
* add multiclass support

* remove debugging print
Sign up for free to 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.

3 participants