Can we use Anchor to explain already built models in our own way?

[Dilini Sewwandi Rajapaksha]

Hi,

First of all, I should say that this is really a great paper. Good luck with you.

My question is, let's assume we have a tabular dataset. We just do encoding categorical variables as pre-processing and make a model using the random forest. And we save that model using .pickle format.

After that can we use Anchor to explain that saved model.

For example, I don't need to use these lines of code when making the model.

explainer = anchor_tabular.AnchorTabularExplainer(dataset.class_names, dataset.feature_names, dataset.data, dataset.categorical_names)
explainer.fit(dataset.train, dataset.labels_train, dataset.validation, dataset.labels_validation)
c = sklearn.ensemble.RandomForestClassifier(n_estimators=50, n_jobs=5)
c.fit(explainer.encoder.transform(dataset.train), dataset.labels_train)

I just need to use,
c = sklearn.ensemble.RandomForestClassifier(n_estimators=50, n_jobs=5)
c.fit(dataset.train, dataset.labels_train)
these two lines when making the model.

Thank you.

[Dilini Sewwandi Rajapaksha]

In LIME, explainer was created, after built the model. But in Anchor, we need to build the explainer before making the global model.

[Marco Tulio Correia Ribeiro]

I'm glad you found the paper interesting.

Yes, you can train the model without encoding, provided that you don't have categorical features (i.e. the explainer will tread all features as ordinal / continuous and discretize them). You don't need to initialize the explainer before training the model, I just did it in that order because I wanted to use the explainer's encoder to encode the data before training the model. Note that the init function of the explainer does not depend on the model.