|
Lecture slides are posted here as the course progresses, along with notes on each lecture.
Note that some slides are password protected because they contain copyrighted figures. These should be readable from any University of Edinburgh machine. To read them from an external machine such as a home system, the username and password can be found in /group/teaching/cnv/README on a UE Informatics DICE machine.
Handing out introductory slides and giving overview of course, focusing on why vision is an important topic, why computational modeling can be useful, and what makes a particular type of computational model appropriate for a given use.
Note that this class differs from NIP (Neural Information Processing) in being much more qualitative, with very little mathematical work required, and by providing extensive background material on vision. It differs from NC by focusing on large numbers of units organized into topographic maps, rather than on more detailed study of individual neurons. It differs from CCN by being focused on specific results from the neuroscience of vision and on models grounded on specific visual areas and circuits within them.
Required reading for next class:
Chapter 1 of the CMVC text, plus sections 2-4 of Report
of the 1st INCF Workshop on Large-scale Modeling of the Nervous
System. Read the
/group/teaching/cnv/README
file mentioned above for information about obtaining the CMVC book.
Beginning review of biological data about the visual system. Covering image formation, the gross anatomy of the visual sysem, and the structure and function of the retina.
Chapter 2 of the text is assigned as background reading.
As we work through this background material, you may find that this article is helpful for explaining any topic that I cover too quickly or that you want to follow up on:
Crick, F. and Asanuma, C. (1986) Certain aspects of the anatomy and physiology of the cerebral cortex. In J. L. McClelland and D. E. Rumelhart (eds.), Parallel Distributed Processing: Explorations in the Microstructure of Cognition, vol. II, chapter 20, pp. 333-371. MIT Press.
The Crick article covers most of the basic concepts in neuroscience
from a modelling perspective, and goes into a lot more detail
about molecular, cellular, and anatomical concepts than we will
discuss in this course. The material in this article is
not examinable, but may be helpful for anyone who does
not have a prior background in this area. There are also a lot
of other basic introductions to neuroscience available; I mention
this one only because it is explicitly written from a modelling
perspective.
Continuing review of the visual system, focusing on cell response types in the retina, LGN, and V1.
Suggested background reading: the vision chapter(s) of any
neuroscience textbook, e.g. Bear, Connors, and Paradiso, Neuroscience:
Exploring the Brain, or Kandel, Schwartz, and Jessell,
Principles of Neural Science. But this is just for
background and more information; as with the basic neuroanatomy
reading above, none of it is required.
Continuing review of the visual system, focusing on feature maps in V1.
Required reading:
Hubel, D. H. and Wiesel, T. (1962). Receptive fields, binocular interaction, and functional architecture in the cat's visual cortex, J. Physiol. 160: 106-154.
Note that this is a long article, and you will not be examined
on any of the details of its contents. So feel free to skim it
at whatever level you prefer. Even so, this is the first
important study of the electrophysiological properties of V1
neurons, and it is well worth reading. Other related
papers on monkey cortex from the same authors can be found on the
Background readings page, if you
are interested.
Continuing review of the visual system, focusing on lateral interactions, feedback, and higher visual areas.
Required reading:
Blasdel, Gary G. (1992). Orientation selectivity, preference, and continuity in monkey striate cortex, J. Neuroscience 12: 3139--3161.
Again, this is a long and very detailed article, and so feel
free to skim it looking for the high points. Even so, it is
an excellent way to understand how optical imaging experiments
are done, the types of analyses that can be done on cortical
maps, etc.
Completing review of the visual system, focusing on development.
Highly suggested reading:
von Melchner, L., Pallas, S.L., and Sur, M. (2000). Visual behaviour mediated by retinal projections directed to the auditory pathway. Nature 404: 871--876.
An earlier review paper might also be good for background:
Sur, M., Pallas, S.L., and Roe, A.W. (1990). Cross-modal plasticity in cortical development: Differentiation and specification of sensory neocortex. Trends in Neurosciences 13: 227--233.
In answer to a question asked in class, a
1993 J. Comparative Neurology paper from the same group
argues that connections from A1 to other cortical areas are
not significantly modified in the rewired ferrets,
suggesting that their behavioral performance is not due to A1
just linking up with the rest of the cortical visual stream.
Watching movies of visual neurophysiology (Hubel and Wiesel),
neural growth cones and axon pathfinding, CNS development, and
LGN activity decoding. Beginning review of modeling approaches
for computational neuroscience of vision, focusing on
non-developmental models for early visual areas.
Continuing review of modeling approaches, focusing on V1 cell models and SOM.
Assigned chapter 3 of the text as background reading.
Continuing review of modeling approaches, focusing on
SOM-based model of learning retinotopy. Discussing first part of
assignment.
Introducing the LISSOM model as a more biologically realistic
but closely related way to develop maps, focusing on a simple
retinotopy model.
Introducing the LISSOM model of orientation maps.
Section 5.3 of the text is assigned as background reading.
Meeting in Appleton Tower Computer Lab South on Level 5,
to choose pairs and work on the first assignment. I'll be
answering questions, and helping everyone find a partner.
Continuing with LISSOM model of orientation maps. Discussing additional types of analyses, plus mechanisms for working with large-scale images: scaling models to larger areas and densities, and contrast-gain control via afferent normalization.
Background reading: Chapter 8 starting with
section 8.2.3, and chapter 15 through 15.2.3 (only skimming
necessary).
Continuing with LISSOM model of orientation maps, focusing on
the function and behavior of the map for real images, in order to
determine which properties are realistic and meaningful, and
which are limitations of the model (and how those can be
improved).
Background reading: chapter 9.
Discussing what visual features would be useful to measure besides orientation, focusing on properties that can be detected reliably through a small circular aperture with a limited resolution. Introducing LISSOM models of ocular dominance and joint models of orientation and ocular dominance.
Background reading: rest of chapter 5.
Discussing LISSOM models of ocular dominance and joint ocular dominance and orientation.
Discussing LISSOM models of motion direction and joint ocular dominance, orientation, and motion direction.
Discussing models of adult visual function, focusing on surround modulation and aftereffects.
Background reading: Chapter 7.
Models of areas beyond V1, and higher-level visual capabilities.
Background reading: Chapters 16, 17, 18.
In AT-5.01 computer lab (NOT the South lab as previously theorized).
Last updated: 2009/01/11 23:39:03
|
Informatics Forum, 10 Crichton Street, Edinburgh, EH8 9AB, Scotland, UK
Tel: +44 131 651 5661, Fax: +44 131 651 1426, E-mail: school-office@inf.ed.ac.uk Please contact our webadmin with any comments or corrections. Logging and Cookies Unless explicitly stated otherwise, all material is copyright © The University of Edinburgh |
