30 Sep 2015
School
Department
Course
Professor
15-05-27
1
LECTURE 6
CHAPTER 6
Generalization and
Discrimination Learning
May 28, 2015
Outline: Generalization and Discrimination
Learning
• Behavioral processes
• When Similar Stimuli Predict Similar Consequences
• When Similar Stimuli Predict Different Consequences
• When Dissimilar Stimuli Predict the Same Consequence
• Brain substrates
• Cortical Representations and Generalization
• Generalization and the Hippocampal Region
Using the past to cope with the present
• Need to respond to stimuli that are completely novel
• Rather than respond randomly, use prior experience with
stimuli we have encountered.
Present Decision
Memory of Past
?
Using the past to cope with the present
Two basic ways of using these past experiences:
1. Generalization: transferring past experiences to new
situations
2. Discrimination: the perception of differences between stimuli
Present Decision
?
=
=
Generalize
Discriminate
Prediction
MT: 47 mc and 3 short answer
15-05-27
2
Using the past to cope with the present
• The right balance:
• Specificity: applying the past narrowly; discriminating
• Generality: applying the past broadly; generalizing
Present Decision
?
=
=
Generalize
Discriminate
Prediction
Generalization vs. Discrimination
Generalization in Learning
• Sensitization: Chapter 3
• Classical conditioning: Chapter 4
• Operant conditioning: Chapter 5
Similar Stimuli Can Predict Similar Consequences
• Training: Yellow light ! Peck ! Food Reinforcement
• Test: presented with novel lights of different colors
• Rate of responding recorded for training and test
15-05-27
3
9
Similar Stimuli Can Predict Similar Consequences
• Generalization gradient—graph showing how physical
changes in stimuli correspond to behavioral response
changes.
• Responding is highest for physically similar stimuli
• Responding is less for less similar stimuli
10
Similar Stimuli Can Predict Similar Consequences
• Represents a default balance between generalization and
discrimination
• Default balance can be modified by experience
Generalization
Discrimination
11
The Challenge of Incorporating Similarity into
Learning Models
• Animals clearly respond to stimuli based on similarity with
past experience
• How can “similarity” be captured in a formal model of
learning?
• Depends on how stimuli are represented
• 4 different systems for representation
• Based on the idea of network models
Network Models
• Stimulus is represented by a set of nodes or stimulus
elements
• EX: Connectionist model
- physical changes to stimuli corresponds to the behavioural changes
- higher amount of responding to physically similar stimuli
- responding is less for less similar stimuli