engleski

How memory colors effect arises in a recurrent neural network?

When observers adjust the colors of diagnostically colored objects to a subjective grayscale they tend to select a shade of gray that contains a small amount of complement color. For example, gray banana is typically adjusted to a slightly bluish gray. This is known as a memory colors effect. Here, we showed how this effect is explained by a neurocomputational model of color perception based on the adaptive resonance theory. The effect is attributed to the adaptation or habituation in a feedback pathway from color categories to the color-opponent representation. This adaptation creates small imbalance in synaptic connectivity that bias competition between color categories towards selection of complement color. We conclude that it is possible to observe memory color effect in a computational model designed to achieve stable color categorization.

memory color effect, recurrent neural network, color perception

nije evidentirano