I would like to know how can the Universal Recommendor's model be trained incrementally. I am not exactly looking for online learning but would like to know if we can train the existing periodically only with the new data. For example, I have trained the model on 10L events and now the application generates 1000 more, do I have to train the model all over again or can I train it only for the 1000 new events? Please help.
I am new to ML and could not find any reference to the above anywhere. Any help would be appreciated. Regards, Pulkit
