GenoLearn Evaluate

Users may wish to evaluate their trained model on a different part of their dataset. Suppose the user would like to evaluate their model on data points that have not been labelled in the metadata csv. Upon executing the evaluate command, the user is prompted to select a train subdirectory and then is prompted for

output filename :
group values*   :

where

• output filename is, as the name suggests, the output filename such that <output filename>.csv and <output filename>.npz will be generated within a subdirectory evaluate within their chosen train subdirectory.

• group values* is a comma seperated string of group values (valid group values will be shown). Only data points corresponding to the group values* will be evaluated in the output files.

Upon a successful execution of the evaluate command, the following should be generated within the user selected train subdirectory

evaluate
├── <output>.csv
└── full-<output>.csv

where both files contain predictions on instances associated with the entered group values*. The first csv corresponds to the model in the parent directory trained only on the train dataset whilst the csv starting with full- corresponds to the model in the parent directory trained on both the train and validation datasets.