parrt
Support for tensorflow - decision forest
https://blog.tensorflow.org/2021/05/introducing-tensorflow-decision-forests.html
Implemented the basic necessary methods to be able to initialize a shadow tree without errors :) . Still work to be done to add the first visualisations
By the way I think in your examples you tend to pass the shadow tree around but nobody knows what that is so I think our example should avoid that unless you are specifically doing something for efficiency reasons.
By examples you mean this screenshots or visualisations from the notebook ?
In the notebook I initialize the shadow tree like this :
tfdf_shadow = ShadowTFDFTree(model, tree_index=0, x_data=train[features], y_data=train[target], feature_names=features, target_name=target, class_names=[0, 1])
and send it to the visualisations instead of using the other method of sending a long list of parameters for each.
I just meant no one knows what a shadow tree is. Why expose that in normal circumstances to the user?
Indeed, if somebody is looking to the above screenshoot, they wouldn't know what a tfdf_shadow is.
But if the user is looking through the notebook visualisations, he/she will have access to the initialization step for tfdf_shadow.
True, but I just want to make sure we show the simplest possible interface for most people. Those were efficiency issues can dig deeper.