EPHE 245 Lecture Notes - Lecture 3: Hebbian Theory, Classical Conditioning, Reinforcement Learning
Document Summary
E(cid:454)a(cid:373) (cid:395)"s: explain how we learn from prediction errors, pe computed all the time, a thing that you are always doing automatically, explain the difference between reinforcement learning, observational learning, and supervised learning. Hebbian learning: means what is going on the neural level, neurons that wire together fire together. Talk a(cid:271)out the st(cid:396)e(cid:374)gth of a (cid:272)o(cid:374)(cid:374)e(cid:272)tio(cid:374) i(cid:374) te(cid:396)(cid:373)s of (cid:862)(cid:449)eight(cid:863) o(cid:396) a (cid:862)(cid:448)alue(cid:863) (cid:449)hi(cid:272)h (cid:373)ea(cid:374)s ho(cid:449) st(cid:396)o(cid:374)g a connection is. Yellow arrow is some other influence on the signal, could be fb (positive fb would strengthen connection, negative fb would weaken it) This is a form of learning that we can observe: change in behavior that is relatively long lasting, qualifies as a form of learning. Developed the rescorla wagner learning rules. Pe: when the outcome is different than the expectation. Vreward may be different depending on how much you like that juice/reward. Learning = values associated with stimuli change, value when taking a certain action change.