Adding a note for future consideration: one reason Causal Survival Forest can be time consuming on large data is fitting (predicting) survival and censoring curves.
A big time chunk is spent in predicting the survival curves via the DefaultPredictionStrategy. We were aware of this when making it, just documenting it here in case we wish to revisit it in the future, it would be possible to speed up CSF by
having survival forest use OptimizedPredictionStrategy at a higher memory cost
have survival forest use some hash table other than the std library that is optimized for dense data
Adding a note for future consideration: one reason Causal Survival Forest can be time consuming on large data is fitting (predicting) survival and censoring curves.
A big time chunk is spent in predicting the survival curves via the DefaultPredictionStrategy. We were aware of this when making it, just documenting it here in case we wish to revisit it in the future, it would be possible to speed up CSF by
(#652)