Transcript SemDevDegenGradVis09

Development and Disintegration of
Conceptual Knowledge:
A Parallel-Distributed Processing
Approach
Jay McClelland
Department of Psychology and
Center for Mind, Brain, and Computation
Stanford University
Some Phenomena in Development
• Progressive differentiation of concepts
• Overgeneralization in naming
• Illusory correlations
Overgeneralization in Naming and
“Illusory Correlations”
• Children often over-generalize object names,
e.g. calling most animals “dog”.
• Rochel Gelman found that many children think
that all animals have feet.
– Even animals that look like small furry balls
and don’t seem to have any feet at all.
Semantic Dementia
• Patients typically in 50s
present with difficulty naming
familiar objects.
• Intact executive functions,
normal memory for unfamiliar
stimuli (novel faces, random
shapes), fluent speech,
spared ability in abstract tests
such as Raven’s matrices.
• Patients have increasingly
severe deficits in tests of
semantic knowledge both
from words and pictures.
• Disease also progresses, from
the temporal pole to
encompass large parts of
anterior and lateral temporal
neocortex.
language
Some Phenomena in Semantic
Dementia
• Progressive loss of conceptual differentiation
• Overgeneralization of frequent names
• Illusory correlations
Picture naming
and drawing in
Sem. Demantia
Parallel Distributed Processing
Approach to Semantic Cognition
• Representation is a pattern
of activation distributed
over neurons within and
across brain areas.
• Bidirectional propagation
of activation underlies the
ability to bring these
representations to mind
from given inputs.
• The knowledge underlying
propagation of activation is
in the connections.
language
A Principle of Learning and
Representation Captured in PDP models
• Learning and representation are sensitive to
coherent covariation of properties across
experiences.
What is Coherent Covariation?
• The tendency of properties of objects to cooccur in clusters.
• e.g.
– Has wings
– Can fly
– Is light
• Or
– Has roots
– Has rigid cell walls
– Can grow tall
The Rumelhart Model
Target output for ‘robin can’ input
Early
Later
Later
Still
E
x
p
e
r
i
e
n
c
e
Overgeneralization of Frequent
Names to Similar Objects
“goat”
“tree”
“dog”
Errors in Naming for As a Function of Severity
Patient Data
Simulation Results
omissions
within categ.
superord.
Severity of Dementia
Fraction of Neurons Destroyed
Ongoing Projects in the Lab
• Semantic Cognition:
–
–
–
–
Domain generality (Tim Rogers)
Bayesian vs. Connectionist Approaches
Relation of semantic and linguistic knowledge (Katia Dilkina)
Integration of semantic and episodic memory
(Cynthia Hendersion)
• Explicit and implicit processes in causal reasoning and
contingency learning (Daniel Sternberg)
• Where do biases and stereotypes come from?
(Jeremy Glick)
• How do brain areas work together when we think,
perceive and remember? (Cynthia Henderson)
Collaborations at Stanford
• Paul Thibadeau and Lera Boreditsky
– How we share knowledge across content domains
• Analogy, metaphor, and cross-domain knowledge sharing
• Juan Gao (post-doc) and Bill Newsome
– Dynamics of decision making; continuous representation of
decision state
• Dharshan Kumaran (post-doc) and Anthony Wagner
– Neural basis of new semantic learning
• You and {…}
– ???