Demystifying the Brain by V. Srinivasa Chakravarthy

Author:V. Srinivasa Chakravarthy
Language: eng
Format: epub
ISBN: 9789811333200
Publisher: Springer Singapore

A large number of real-world complex patterns are similarly reduced. The resulting patterns are definitely more complex than oriented lines that are V1’s favorites, but less complex than the real-world objects from which they were derived. These patterns seem to be some sort of archetypes of the visual world by which the visual brain breaks up and represents visual objects. A few logical questions are in order: Is there a map of complex shapes in IT? Is there columnar structure? Analogous to orientation columns in V1, the researchers found columnar structure in IT. Neurons at various depths at the same point, or nearby neurons within a diameter of about 400 μm, responded to similar objects.

Further exploration of the responses of IT neurons to complex visual objects revealed other elegant aspects of how real-world objects are represented in this part of the brain. Until now, we have been depicting visual objects as complex two-dimensional patterns. But real-world objects are three dimensional. They must be associated with not one but a whole continuum of two-dimensional patterns, where each pattern corresponds to a different view of the same object. Therefore, a neuron that claims to recognize a three-dimensional object must respond not just a single view of the object, but several, if not all.

The appearance of an object changes as our spatial relationship with the object varies. Objects seem to grow larger as we move closer to them. Change in appearance is even more complicated as we move around an object sweeping a large angle. But if the angle swept is relatively small, the object appears to shrink or expand in the direction of our motion. Change in distance changes the size of the appearance of the object. On the other hand, if our motion is in horizontal direction, then the shrinkage or expansion also occurs in horizontal direction, leaving the height invariant. In other words, such a motion of the observer changes the aspect ratio of the appearance of the object. An ability to recognize three-dimensional objects implies a certain level of invariance in recognition with respect to changes in scale and aspect ratio. Such invariance was indeed discovered in IT neuronal responses by Hosein Esteky and Keiji Tanaka. When two-dimensional patterns were presented at different sizes, neurons were found to respond significantly to a range of scales spanning over two octaves. Similarly, when planar patterns were presented at different aspect ratios, neurons were found to respond with greater than 50% of the highest response even over a variation of thee octaves in aspect ratio.

Just as other visual maps encountered so far in our discussion, the “object maps” of IT area are not cast in stone and can be altered by changing stimulus conditions. In order to study the effect of training on IT neuronal responses, Tanaka and colleagues trained monkeys to recognize simple, artificial, and geometric patterns by giving the animals juice rewards whenever they recognized the patterns accurately. The training went on for 3 to 4 months. After training several

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.