Does the models support update ?

[dangmanhtruong1995]

Hi, may I ask, if after training I have another additional data point then does the model support update or do I have to retrain from the beginning ? I checked the doc and there was update_training_values() but it does not appear to add additional data instead it resets the train set. Thanks a lot.

[Paul-Saves]

Model predictions are space-dependent. Adding a point to the model means recomputing the model all over the space. When using hyperparameters, we can't know beforehand where the model will be impacted by adding a point neither what the influence will be. The only way to know that is to build a new model and reset the training set. This is related to leave-one-out error.
This is why for optimization (see applications/ego), we are building a new model for every point we add.