Self-organization of connection patterns within brain areas of animals begins prenatally, and has been shown to depend on internally generated patterns of neural activity. Such activity is genetically controlled and has been proposed to give the neural system an appropriate bias so that it can learn reliably from complex environmental stimuli. This paper demonstrates this idea computationally. A competitive learning network is trained with hand-designed patterns during a prenatal developmental phase, and its classification performance in a line categorization task is significantly affected as a result. Plotting and analyzing the network weights during various stages of the learning process reveals the complex dynamics through which the bias is established, and suggests that evolution might be necessary to discover the appropriate pattern generators automatically. This approach is expected to be useful in building complex artificial systems, such as the learning system of a robot with uninterpreted sensors and effectors.
|Title of host publication||Proceedings of the Fifth International Conference on Development and Learning (ICDL-2006)|
|Publication status||Published - 2006|