search

Indicia-Team / warehouse

GNU General Public License v3.0
5 stars 3 forks source link

Sample level training #228

Open andrewvanbreda opened 7 years ago

andrewvanbreda commented 7 years ago

I noticed a problem when working on a project, am not sure how this will be resolved. The design of the system currently is that training records are set at the occurrence level, but this is theoretically an issue if we are interested in empty samples. For a project test I created some occurrence training records, but the report was showing this a "proper" sample/visit which just didn't have any occurrences. This situation of not sightings was of interest to the project but was not correct in this scenario as it was only like that due to the way training records work. Would it be better to have training set at a sample level as well?

johnvanbreda commented 7 years ago

Yes, or it might make sense to have the training flag at sample level instead as discussed the other day - though this would be more work.

kazlauskis commented 7 years ago

Hi, I also have this requirement where NPMS App surveys can have no occurrences and so no way to send an empty training sample.

Also, I feel this makes me more sense to set a training on a parent level and treating its children models as training rather than the opposite, ie. I wouldn't consider sending one of the occurrences within a sample as training while other being valid models.

andrewvanbreda commented 7 years ago

I have just thought of something Karolis, the reason why the NPMS site probably isn't working properly with training records is that the grids display samples (NPMS visits) not occurrences. I will email Oli about this as I don't think we will be able to get round this. Certainly not easily enough to be worthwhile doing unless someone changes things so that training is associated with the sample. I think for now your NPMS test data will just have to be created normally and then deleted.