Detailed Course Description

Computational models of visual perception, attention, and awareness

                                         Michael C. Mozer

(homepage)

University of Colorado

Department of Computer Science and Institute of Cognitive Science

      This course will describe computational models as they apply to psychological

and neuropsychological phenomena in visual perception, spatial attention,

and awareness.  The lectures will emphasize connectionist modeling

techniques but will also utilize alternative probabilistic formalisms

including Bayesian belief networks.  A unifying theme of the course will

be to view perceptual systems as producing well-formed interpretations of

the noisy visual stimulation, and to conceptualize cognition in terms of

the transmission of information among cortical systems.  The course will

address data from the cognitive sciences including:  perceptual illusions,

attentional cueing effects, object-based attentional effects, word reading,

the neuropsychological phenomena of unilateral neglect and optic aphasia,

perceptually-based problem solving, and long-term priming -- both supraliminal

and subliminal.

      1. Modeling visual perception and spatial attention

- transformation invariant recognition

- early and late attentional selection

- word reading

- binding errors and perceptual illusions

Reading: Mozer, M. C. & Sitton, M. S. (1998).
Computational modeling of
spatial attention.  In H. Pashler (Ed.), “Attention” (pp. 341-393).

London:  UCL Press. 

      2. Modeling deficits of attention

- location-based and object-based modes of attention

- unilateral neglect

- frames of reference in perception

Reading: Mozer, M. C. (2002).  
Frames of reference in unilateral neglect: 
A computational perspective.  “Psychological Review”, 109,

156-185.

      3. Modeling deficits of visual perception

- optic aphasia

- content-specific neglect

- prosopagnosia

      Reading: Sitton, M., Mozer, M. C., & Farah, M. J. (2000)
 Superadditive effects of multiple lesions in a connectionist architecture:
Implications for the neuropsychology of optic aphasia.

“Psychological Review”, 107, 709-734.

Reading: Farah, M. J., O'Reilly, R. C., & Vecera, S. P. (1993).

Dissociated overt and covert recognition as an emergent property

of a lesioned neural network.  “Psychological Review”, 100,

571-588.

      Reading: Zemel, R. S., & Mozer, M. C. (2001).  
Localist attractornetworks.  “Neural Computation”, 13, 1045-1064.

      4. Modeling learning of perceptual skills

- repetition priming

- bias and sensitivity effects in priming

- repetition suppression

      Reading:  Mozer, M. C., Colagrosso, M. D., & Huber, D. H. (2002).
A rational analysis of cognitive control in a speeded

discrimination task .  To appear in T. Dietterich, S. Becker,

& Z. Chahramani (Eds.), “Advances in Neural Information Processing

Systems 14”.  Cambridge, MA:  MIT Press. 

      Reading:Mozer, M. C., Colagrosso, M.D., Huber, D. E. (2002).
 Mechanisms of skill refinement: A model of long-term repetiion priming

      Reading:  Ratcliffe, R., & McKoon, G. (1997).   
A counter model for implicit priming in perceptual word identification.  

“Psychological Review”, 104, 319-343.

      Reading:  Rainer, G., & Miller, E. K. (2000).   
Effects of visual experience on the representation of objects in the prefrontal cortex.

“Neuron”, 27, 179-189.

      5. Modeling perceptual awareness

- subliminal priming paradigms

- stability theory of awareness

- necessary and sufficient conditions for awareness

      Reading:  Mathis, D. A., & Mozer, M. C. (1996).  
Conscious and unconscious perception:  A computational theory  

In G. Cottrell (Ed.),“Proceedings of the Eighteenth Annual Conference of the

Cognitive Science Society” (pp. 324-328).  Hillsdale, NJ:Erlbaum.

Reading:  Kanwisher, N. (2001 ).

 Neural events and perceptual awareness.“Cognition”, 79, 89-113.

      Small groups

      Participants will explore several of the models in greater detail, focusing

on the mechanics and mathematics of the models.  The session will include

computer simulation demonstrations.

(1) a connectionist model of object recognition and spatial attention (MORSEL)

(2) a connectionist model of object-based attention (MAGIC)

(3) localist attractor networks

(4) belief networks

(5) a model of anagram problem solving

      Assessment

      I propose that students who desire credit write two brief papers:  one

reviewing a subset of literature discussed in the course, and one discussing

how the modeling ideas presented in the course could be applied to other

experimental data, either psychological or neurobiological.

     Michael Mozer

      Michael Mozer received a Ph.D. in Psychology and Cognitive Science at the

University of California at San Diego in 1987.  Following a postdoctoral

fellowship with Geoffrey Hinton at the University of Toronto, he joined

the faculty at the University of Colorado at Boulder and is presently

a Professor in the Institute of Cognitive Science and the Department

of Computer Science.  He served as Chief Scientist at Athene Software

(1997-2001), Director of Machine Learning Technologies for Sensory,

Inc. (1993-present), and on the technical advisory boards of Green Planet

Software and AnswerOn (2001-present).  He is the recipient of a 1990 NSF

Presidential Young Investigator Award, and has over 80 publication in neural

networks and machine learning.  He is presently on the Governing Board

of the Cognitive Science Society.  He has been an organizer of the Neural

Information Processing Systems (NIPS) conference, serving as its program chair

and general chair, and is presently a member of the NIPS Foundation Board.

His research interests include computational models of human perception,

attention, memory, and awareness, and the application of machine learning

prediction and control techniques to intelligent environments, speech and

language recognition systems, and customer relationship management.

Power Point Presentations:

 | Lecture 1 | Lecture 2 | Lecture 3 | Lecture 4 | Lecture 5 |