Classical Conditioning I: Prediction learning, Exercises of Decision Making

how did you solve the credit assignment problem? What should you learn from interaction with the world? 1.what is going to happen (prediction learning).

Typology: Exercises

2022/2023

Uploaded on 02/28/2023

picoo
picoo 🇮🇳

4.5

(13)

235 documents

1 / 9

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
Classical Conditioning I:
Prediction learning
PSY/NEU338: Animal learning and decision making:
Psychological, computational and neural perspectives
Decision making
2
Why is this hard?
Reward/punishment may be delayed
Outcomes may depend on a series of actions
󲰛 “credit assignment problem” (Sutton, 1978)
pf3
pf4
pf5
pf8
pf9

Partial preview of the text

Download Classical Conditioning I: Prediction learning and more Exercises Decision Making in PDF only on Docsity!

Classical Conditioning I:

Prediction learning

PSY/NEU338: Animal learning and decision making: Psychological, computational and neural perspectives

Decision making

Why is this hard?

  • Reward/punishment may be delayed
  • Outcomes may depend on a series of actions “credit assignment problem” (Sutton, 1978)

another example:

3 how did you solve the credit assignment problem?

What should you learn

from interaction with the world?

1.what is going to happen (prediction learning) 2.what to do about it (action learning)

example 1I: fear conditioning

(conditioned suppression)

7 Habituation (tone) Conditioning (tone+shock) Extinction (tone) CS: Tone, 30 sec US: Shock, 0.5 sec CR: Freezing (ITI = 4 min) Quirk Lab, University of Puerto Rico

example 1I: fear conditioning

(conditioned suppression)

  • Pavlov called the US a^ “reinforcer” What does that mean?
  • Purely^ operational^ definition^ (makes no assumptions regarding affective components)
  • Acquisition
  • Extinction
  • Predictions are: 1) shaped by^ experience
    1. revealed by behavior

some non-trivial terminology

9 = reinforcer

what makes conditioning

Pavlovian?

procedurally: Pavlovian/classical conditioning is a learning situation in which the reinforcer does not depend on the animal’s response from the animal’s point of view: the conditioned response is unaviodable , like a reflex , not utilitarian or flexible; direct result of a prediction (e.g., Hershberger (1986) - An approach through the looking glass)

  • Different USs are more easily conditioned to certain CSs (eg. Garcia and Koelling’s “noisy water” experiment, with shock US versus LiCl US)
    • also depends on the animal species: pigeons associate color with illness, rats - flavor
    • evolutionarily adaptive constraint
  • The CR is^ mostly similar^ to the UR
    • does not have to be similar: freezing versus jumping as a response to shock, salivating versus biting as a response to food

CS-US-CR compatibility

13

“stimulus substitution”

outline

PART 1 - Basics of classical conditioning PART II - Some challenging results PART III - A theory (model) 15

back to basic classical

conditioning

What is the proper control experiment?