search

scikit-learn-contrib / metric-learn

Metric learning algorithms in Python
http://contrib.scikit-learn.org/metric-learn/
MIT License
1.4k stars 234 forks source link

Combination of supervised and weakly-supervised data #233

Open bellet opened 5 years ago

bellet commented 5 years ago

We could easily allow users to fit weakly-supervised algorithms on a combination of label supervision (from which we generate constraints as in supervised versions) and additional weak supervision specified by the user

hansen7 commented 5 years ago

cool! actually the semi-supervised means to use unlabelled(or without relative comparison etc.) data to construct the loss terms, such as entropy, but how to build up the optimisation process based on this arbitrary loss/constraint?

bellet commented 4 years ago

cool! actually the semi-supervised means to use unlabelled(or without relative comparison etc.) data to construct the loss terms, such as entropy, but how to build up the optimisation process based on this arbitrary loss/constraint?

Yep, I have rephrased the issue