I am working on an online stream data. I have a database that contains image-class tuples. I have trained a model that gives good result but I get new data continiuesly. What are the methods to handle these new samples for improving the model? Is there a good deep learning architecture for this? Bayesian NN’s looks difficult for me.