COGS 110 Lecture Notes - Lecture 16: Correlation Does Not Imply Causation, Reinforcement Learning
Document Summary
Cogs 110 - lecture 16 - decision making. Our brain"s job is to make decisions maximizing reinforcement per unit of time. Rescorla wagner model good, but may be hard to apply sometimes. In real world, not necessarily getting reward all the time. Rescorla wagner model works if life divisible into discrete trials on which there is always reward (either positive or negative) The sooner you get reward, the better. V = s ( r + v(t+1) - v(t) ) V(t+1) - estimated value at time (t+1) V(t) - estimated value at time t. First, don"t know if should go or stay home & study. Place turns out to be terrible, but haven"t entered party yet. Changes idea of how good it is to be invited to party. Changes expectations of what it"s like ot be standing outside in sketchy place. Makes being out front of sketchy place not so bad.