M2 Lecture 13 Flashcards
define Categories
The systematic grouping of instances that are similar
Categories: The systematic grouping of instances that are similar ¤ Different theories on how this grouping occurs
¤ Theories tend to follow what idea
the idea of cognitive economy
what is the idea of cognitive economy
¤ A balance between simplification and differentiation with categorization ¤ Use the fewest ‘bits’ of information to store a category but still ensure a
category can distinguish between things
define Concepts
The knowledge associated with a category
Conceptual knowledge is used to:
¤ Make predictions about our environment ¤ Identify new instances of categories
¤ If it walks like a duck, and talks like a duck …
what is the function of categorizing
allows is to do “the right thing with the right kind of thing
¤Provides material for thoughts
¤The basis of communication
¤Language categories
¤Generalizations to guides behavior and make predictions
¤Stereotypes as social concepts gone wrong
Categories are boundaries between what
collections of instances/members
what are the types of categories
¤ Between-category separation
¤ Within-category compression
what is Within-category compression
¤ Members of the same category look more similar
what is ¤ Between-category separation
¤ Members of different categories
look more distinct
How do we learn concepts?
Rule based approaches
Probability based approaches
what are the Rule based approaches
Concepts are collections of necessary and sufficient features
related to a category
what are the Probability based approaches
¤Concepts and categories are formed through experience ¤The prototype theory
according to a rule based approach, Concepts are made up of what
attributes
what are attributes
Attributes are features of a concept
¤ These can take on different values
¤ Attributes are expressed as different values by category members
according to a rule based approach, Learning a concept involves what
forming rules and testing them
¤ We form hypotheses about the combination of attributes that define a concept
¤ We actively test and refine these hypotheses
what are the Types of concepts
¤ Conjunctive concept
¤ Disjunctive concept
¤ Relational concept
what is Conjunctive concept
¤ Concepts that contain conjunctions of attributes (AND) ¤ A Mother is defined as a female AND having a child
what is Disjunctive concept
¤ Concepts where two or more possible sets of attributes (Either-or
concepts)
¤ Fame can be defined in different ways
what is Relational concept
¤ Concepts defined by a relation between attributes ¤ Marriage is defined by the link between two people
what are the different Rule-based instances
positive
negative
abstraction
what is Positive instance:
An example of a particular concept
¤ Criterial attribute
¤ An attribute that must be present for a positive instance
¤ Wings of a bird
what is Negative instance:
An example that does NOT contain the right attributes of a particular concept
what is Abstraction
What we do to determine which attributes to include and exclude when forming concept rules
¤ Most typical of the concept
explain the experiment that tested how we learn concepts
¤Tested how we learn conjunctive concepts
¤ Two attributes must be present
¤ E.g. A card with a black (AND) square
¤Reception task
¤ Participants are shown cards (instance) from a deck in a pre-arranged order
¤ They indicate if they think a card is part of the concept
¤ Experimenter tells them if they are correct
¤Selection task
¤ Participants pick a card (instance) from the deck
¤ Experimenter tells them if it is a positive instance of the concept they are to learn
Selection tasks: Forming rules what is this
Form rules of a concept to guide how to pick instances
what is Conservative focusing
¤ When you encounter a positive instance
¤ Focus on one attribute of this instance to create a rule
¤ ‘Test’ this rule on new instances
¤ Select instances that vary only on this attribute
what is Focus gambling
¤ When you encounter a positive instance
¤ Focuses on all the attributes of this instance to create a
¤ Select new instances that have all these attributes
¤ Quick results if positive instances comes soon after forming rule