Exploring the Potential of Optimal Active Learning via a Non-myopic Oracle Policy.

Christoph Sandrock Marek Herde Daniel Kottke Bernhard Sick

Published in: DS (2023)

Keyphrases

active learning
asymptotically optimal
dynamic programming
expected cost
experimental design
machine learning
learning algorithm
optimal solution
control policy
training set
object oriented
worst case
optimal strategy
reward function
batch mode
selective sampling