The Shifting View of Vision

"I see nobody on the road," said Alice. "I only wish I had such eyes," the King remarked in a fretful tone. "To be able to see Nobody! And at that distance too! Why, it's as much as I can do to see real people, by this light!"
-- Lewis Carroll, Through the Looking Glass

There's more to seeing than meets the eye. Quite a lot more actually. What meets the eye is light waves. Their impact on the retina triggers impulses that travel a pathway along the optic nerve through the thalamus to the back of the brain, where they register on the primary visual cortex. From there, via myriad connections, neuron to neuron, visual impulses reach out to other parts of the brain, and we get an instantaneous mental image of a colorful, three-dimensional world. How does it happen?

The human brain, say scientists who study it, may be the most complex structure in the universe. A convoluted mass of tissue made of a 100 billion interconnected neurons spawns a mental world -- all the processes we call thought, consciousness, creativity, emotion, memory and vision.

Probably the most studied of these is vision, and contrary to long accepted views that brain structure is genetically determined, research since the mid-70s has shown that seeing is a learning process. When a kitten sees only horizontal lines for the first few months of its life, for example, it loses the ability to see vertical lines. These and other experiments suggest that the interconnections between neurons aren't fixed at birth, but evolve depending on visual experience.

"Experiments have shown that much of the structure thought to be innate or hard-wired actually develops during the early days or weeks of life," says Risto Miikkulainen, a computer scientist at the University of Texas, "and even in the adult these structures can reorganize." Using the CRAY T3D at Pittsburgh Supercomputing Center, Miikkulainen and Ph.D. candidate Joseph Sirosh have developed the first computational model to comprehensively mimic this "self-organizing" process.

"We couldn't have done this 10 years ago," says Sirosh, "because the technological capability to simulate these large neural networks didn't exist. You need to model at the scale of tens of thousands of neurons, hundreds of millions of connections. Machines like the T3D are absolutely crucial to this work."

Self-Organization of the Orientation Map
Using oriented light spots as input, LISSOM modeled how orientation preference and neuron-to-neuron connection patterns develop in the cortex. The color of each neuron in the network (192 x 192 neurons), from red to magenta to blue to green, corresponds to its orientation preference from zero to 180 degrees. The small white dots show lateral connections of the neuron marked with a big white dot.

Initially (left) the orientation preferences were random and lateral connections covered almost the entire map. After a series of several thousand inputs, the neurons organize (right) into orientation columns, and lateral connections link areas of similar orientation preference. These patterns agree in key features with maps obtained by experimental imaging.

Researchers: Risto Miikkulainen & Joseph Sirosh, University of Texas.
Hardware: CRAY T3D
Software: LISSOM
Keywords: Vision, sight, eyes, brain, primary visual cortex, visual field, retina, optic nerve, neurons, orientation columns, neuron groupings, orientation preference, self-organization, Hebbian learning, Hebbian algorithm, lateral connections, receptors, neural modeling, cognitive science.

Related Material on the Web:
The University of Texas Neural Nets research group home page
Projects in Scientific Computing, PSC's annual research report.

References, Acknowledgements & Credits