search

MetOffice / XBTs_classification

Project for the classification of eXpendable Bathy Thermographs
BSD 3-Clause "New" or "Revised" License
4 stars 2 forks source link

try additional features to replace country #27

Closed stevehadd closed 4 years ago

stevehadd commented 4 years ago

For some unlabelled profiles, there are no labelled profiles in the dataset from that country, so the classifier has not been trained on any examples frm that country and doesn't know how to deal with such profiles. This is a problem in the code pipeline. For now we will fallback on using iMeta in these cases (#26), but we should explore whether including additional/alternative input features (cruise ID, institute, paltform, lat/long) might give comparable results while being able to handle all the unlabelled data.

stevehadd commented 4 years ago

This has now been demonstrated in notebooks and implemented in the experiment framework.