Found 1 document(s)
Page 1 of 1
-
Learning useful representations from continually changing value targets remains a challenge in deep reinforcement learning. We study the temporal horizon of this supervision: smaller discounts make these targets less sen...