Closed glemaitre closed 1 month ago
The goal is to be consistent with the conventions of scikit-learn where all predict_
methods return datastructures with a first dimension that is aligned with the observations onto which the predictions are made.
Change some API such that the shape of returned arrays are:
predict_proba
is an array of shape (n_samples, n_events)predict_cumulative_incidence
is an array of shape (n_samples, n_events, n_times)predict_survival_function
is an array of shape (n_samples, n_times)