Neural network model for the mechanism of visual pattern recognition

Kunihiko Fukushima

The goal of this study is to understand the mechanism of visual pattern recognition in the brain. We use modeling approach. In other words, we study how to interconnect neurons to synthesize a network with the same functions and abilities as the brain. When synthesizing a model, we try to follow physiological and psychological evidence as faithfully as possible. For parts that are not yet clear, we construct a hypothesis and synthesize a model that follows the hypothesis. We then analyze or simulate the behavior of the model, and compare it with that of the brain. If we find any discrepancy in behavior of the model and the brain, we change the initial hypothesis and modify the model. We repeat this procedure until the model behaves in the same way as the brain.

Fig.1

We especially concentrate upon the mechanism of character and object recognition, where the interaction of bottom-up and top-down signals in hierarchical neural networks plays an important role.

Fig.2

Fig. 2 shows an example of the problem we are now tackling. Partly occluded patterns shown in (a), in which the occluding objects are invisible, are almost illegible, but the patterns in (b), in which the occluding objects are visible, are much easier to read. A neural network model that responds to these stimuli like a human being is an example of the models we are trying to construct.