Post by Paul Khuong

@bob you can do active learning with asynchronous methods (adorn the ids in that list with some weight). Thompson sampling is very robust to a bit of delayed feedback, for example. Context is indeed…

By Paul Khuong

View on We The Citizens