Garrison W. Cottrell

University of California, San Diego

My lab has been involved in building neurocomputational models of face processing for fifteen years now. Even though that means I'm getting to be really ancient now, I still have a few reasonably cogent things to say. Computational models can provide insights into the possible mechanisms of face processing that behavioral, imaging, electrophysiological and single cell recording technologies cannot. In computational models, one can "hold a stick" into a model cell from "birth" on; one can follow possible alternative life experiences for a cell, and one can explore the representations developed in a whole population of cells. In this course, I will review some of issues in face and object processing, review a few simple techniques used in image processing, and show how models can be used to account for a variety of results in face and object processing (but mostly face processing!). I will cover feature extraction, expression recognition and holistic processing, the ideas relevant to the development and role of the Fusiform Face Area and the surrounding controversy, holistic processing, how expertise generalizes, the other race effect,and time permitting, describe some of our future directions in modeling active sampling of images in the form of eye movements.

Lecture 1. Issues and techniques in face and object processing PowerPoint Presentation

Required readings

Palmeri, T. and Gauthier, I (2004) Visual Object UnderstandingNature Reviews Neuroscience 5:291-303.

Riesenhuber, M. and Poggio, T. (1999). Hierarchical models of object recognition in cortex. Nature Neuroscience 2:1019-1025.

Optional readings

Cottrell, G. (1990). Extracting features from faces using compression networks. In Touretzky, D.S., Elman, J.L., Sejnowski, T.J. and Hinton G.E. (Eds.)

Proceedings of the 1990 Connectionist Models Summer School,pp. 328-337. San Mateo: Morgan Kaufmann. (I am getting this scanned in).

Olshausen, B. (2003) Principles of Image Representation in Visual Cortex In The Visual Neurosciences, L.M. Chalupa, J.S. Werner, Eds. MIT Press: 1603-1615.

Lab: Principal components analysis of faces

Lecture 2. Neural Network models of facial expression recognition PowerPoint Presentation

Face Space Modeling: 2D and 3D
Expression recognition via neural networks: The EMPATH model

Required readings

Dailey, Matthew N., Cottrell, Garrison W., Padgett, Curtis, and Ralph Adolphs (2002) EMPATH: A neural network that categorizes facial expressions. Journal of Cognitive Neuroscience 14(8):1158-1173.

Leopold, D., O'Toole, A. J., Vetter, T. & Blanz, V. (2001). Prototype-referenced shape encoding revealed by high-level aftereffects. Nature Neuroscience, 4,, 89-94

Optional readings

Rumelhart, D.E., Hinton, G.E., & Williams, R. (1986). Learning internal representations by error propagation. In D.E. Rumelhart and J.L. McClelland (Eds.) Parallel Distributed Processing:

Explorations in the Microstructure of Cognition, Vol. 1. Cambridge, MA: MIT Press. Ch. 9.

(http://crl.ucsd.edu/~elman/Bulgaria/rumelhart-hinton-williams.pdf)

Lab: perceptron learning; back-propagation learning of facial expressions.

Lecture 3. A model of the development of FFA PowerPoint Presentation

            The Fusiform Face Area
            The expertise hypothesis
            A model of the development of FFA.

Required readings

Kanwisher, N., McDermott, J., & Chun, M. (1997) The Fusiform Face Area: A Module in Human Extrastriate Cortex Specialized for the Perception of Faces. Journal of Neuroscience. 17 4302-4311

Dailey, Matthew N. and Cottrell, Garrison W. (1999) Organization of Face and Object Recognition in Modular Neural Networks. Neural Networks 12(7-8):1053-1074.

Optional readings

none

Lab: More back propagation learning of faces.

Lecture 4. Visual Expertise PowerPoint Presentation

Evidence for and against visual expertise
The visual expertise "mystery"

Required readings

Gauthier, I., Tarr, M.J., Anderson A.W., Skudlarski, P. & Gore, J. C. (1999). Activation of the middle fusiform "face area" increases with expertise in recognizing novel objects. Nature Neuroscience, 2(6): 568-573.

Joyce, C. & Cottrell, G. Solving the Visual Expertise Mystery
In In Connectionist Models of Cognition and Perception II: Proceedings of the Eighth Neural Computation and Psychology Workshop, Howard Bowman and Christophe Labiouse (Eds.), World Scientific.

Optional readings

Kanwisher website: Domain Specificity for Faces versus Expertise: A Critical Look at the Evidence

Tong, Matthew, and Garrison W. Cottrell (2005) Are Greebles special? Or, why the Fusiform Fish Area would be used for sword expertise (if we had one). In Proceedings of the 27th Annual Cognitive Science Conference, La Stresa, Italy. Mahwah: Lawrence Erlbaum.

Tran, Brian, Joyce, Carrie A., and Garrison W. Cottrell (2004) Visual expertise depends on how you slice the space. In Proceedings of the 26th Annual Cognitive Science Conference, Chicago, Illinois. Mahwah: Lawrence Erlbaum.

Lab: If it gets done in time, a visual expertise lab.

Lecture 5. Generalization gradients in expertise PowerPoint Presentation / Presentation 2

            The other race effect
             Owls and Wading birds: Generalization gradients in expertise
             Modeling of these effects.

Required readings

Tanaka JW, Curran T, Sheinberg DL (2005) The training and transfer of real-world perceptual expertise.
Psychol Sci. 2005 Feb;16(2):145-51.

Nguyen, Nam and Garrison W. Cottrell (2005) Owls and Wading Birds: Generalization gradients in expertise In Proceedings of the 27th Annual Cognitive Science Conference, La Stresa, Italy. Mahwah: Lawrence Erlbaum.

Optional readings

Nakayama, K., He, Z., & Shimojo, S. (1995) Visual surface representation: A critical link between lower-level and higher level vision. In Kosslyn, S. & Osherson, D. (eds.) Vision: An Invitation to Cognitive Science MIT Press pg: 1-70

Olshausen, B. & Field, D. (in press) How close are we to understanding V1? Neural Computation

Itti, L. & Baldi, P. (2005). A principled approach to detecting surprising events in video. In: Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

Afternoon section meetings

Afternoon sessions will be used to explore specific topics in depth and to present additional material.

Assignments

Students who take the course for credit will be asked to write a brief (5-7 page) paper that critical reviews one or more of the articles read in class, or to comment on other work that is related to the issues discussed in the class.

Garrison W. Cottrell

Garrison W. Cottrell was born at a very early age. Despite starting life quite young, andbeing brought up by the family collie, Tippy, who herded him around the backyard, he managed to attend Woodstock, While he was not "influenced by the Russians" as his mother believed, while at Cornell, he did spend a great deal of time protesting the war and had his constitutional rights violated by the Nixon administration. Upon graduation from Cornell University in 1972 with a double major in Mathematics and Sociology, Cottrell was surprised to find that the revolution had not occurred. Desperate for something to do, and unable to leave Ithaca, he immediately enrolled in the Teaching Masters program, and eventually obtained an MAT in Mathematics and a permanent teaching certificate for high school math, grades 7-12 in New York State. After years of school bus driving, ice cream scooping, and rough carpentry, he decided to return to the Academy, eventually obtaining a Ph.D. in Computer Science in 1985 from the University of Rochester under James F.Allen.His thesis concerned a connectionist model of word sense disambiguation that accounted for human data on lexical access. He then became a Postdoctoral Researcher with David E. Rumelhart at the Institute of Cognitive Science at UCSD. His work at this time was on image compression using back propagation. In 1987, he joined the Computer Science and Engineering Department at UCSD. He is currently the Director of the Interdisciplinary Ph.D. Program in Cognitive Science at UCSD, as well as the Program in Vision and Learning in Humans and Machines, an NSF-sponsored interdisciplinary training program. Professor Cottrell's main interest is Cognitive Science combined with a Computer Science salary. With his students, he has built working models of cognitive processes and used them to explain psychological or neurological processes. In recent years, he has focused upon face processing, including face recognition, face identification, and facial expression recognition. He has also worked in the areas of modeling psycholinguistic processes, such as language acquisition, reading, and word sense disambiguation. His most well-known work, however, is probably in the area of Connectionist Dog Modeling, or Dognitive Science, as well as his work on the Connectionist Air Guitar. These have been published in the Humour (sic) section of the journal Connection Science.