7 - Conceptual Learning and Semantic Knowledge Flashcards
define typicality
object showing most usual characteristics of a particular type or thing therefore being a good example of that type
why can items be a less typical example even when they show similarities to the prototype
share fewer similarities than other category members
define economy of storage
propositions of many items stored once and apply to anything underneath it
define generalisation
inference of unseen properties about a concept to new members based on what we already know
The defining attributes model is set up how
categories stored as concepts from super to subordinate and categories get narrower in lower levels
why is defining attributes bottom-up
look in the environment and informed by experiences as look at if novel things have certain attributes to determine its category
properties and concepts stored where are more easily accessible
at the bottom so accessed faster
define prototype approach
categories have a central description or prototype standing for the whole category which the concept is represented by
(prototype) category members share what
family resemblance as have common features but look different
(prototype) what is the prototype informed by
experiences and change more w more experience
(prototype) the prototype has what attributes
average and characteristic ones of all things in that category, not defining
(prototype) how are items categorised
if they’re similar enough to the prototype but don’t have to have exact same attributes unlike defining attributes
define graded membership in the prototype approach
members closer to similarity to the prototype are better members of the category than those further away
(prototype) we compare novel things to what
things we have knowledge and representations of already
define the exemplar approach
compare novel thing to set of all known instances of examples of a category that come to mind which we have encountered
what is exemplar based reasoning
categorisation relies on knowledge about specific members rather than the prototype
(exemplar) how do we categorise
based on similarity to exemplars and if similar enough
(exemplar) what information is preserved
how there is variability between things’ attributes
what does the prototype approach do to all instances of something
averages them and excludes info about their variability
why can objects be categorised in the same category after superficial changes
deep properties count more than superficial ones
define connectionist approach
brain is like a computer and deals with input, processing, and outputs and makes mdoels
what is parallel distributed processing
experience-driven network, as initially have no knowledge or connections
(PDP) what is each idea represented by
pattern of activation across the network
(PDP) how can we train the network
present it with experiences based on info in the hierarchy which the network leads to map onto different attributes
