Prediction on held-out data

[willtownes]

Is it possible to make predictions using a fitted MEFISTO model on held-out test data that were not used in training?

[willtownes]

Copying over from a slack thread to hopefully boost visibility. I understand the way to do this is through the entry_point.predict_factor method. When I tried this it throws the following error. It doesn't appear to matter whether I pass in a new set of covariates or just call the method with no arguments. According to the trace the error occurs on this line.

ValueError: all the input array dimensions for the concatenation axis must match exactly, but along dimension 0, the array at index 0 has size 1000 and the array at index 1 has size 2363

For context, 1000 is the number of inducing points and 2363 is the number of observations in the training data.

[bv2]

Hi Will,
thanks for reporting this bug.
This should be fixed now on the dev branch (pip install git+https://github.com/bioFAM/mofapy2@dev) and will be part of the next mofapy2 version.

[khalilouardini]

Hi!

Has this been fixed on the latest version (entry_point.predict_factor on held out data) please?