Accepted:
01 Apr 1996
| |
comp96e3.html | © Copyright April 1996 | |||
| Volume 03 | Received: Accepted: |
01 Apr 1996 01 Apr 1996 |
|||
|
Do Local Neural Circuits Make Good Global Architectures?
David M. Alexander |
|
| Abstract | |
| This paper deals with the evolution of brain complexity. Evolution introduces new behavioural and cognitive properties by increasing the structural complexity of brains. Viewed in terms of evolutionary stages, there are two ways or directions in which neural structures can become more complex: reaching up and reaching down. Reaching up involves replication of existing neural circuits and co-ordinating them as a larger structure. Reaching down involves local enlargement of circuits, while maintaining the old circuitry as the co-ordinating architecture. Previous experimentation by Walker and Alexander (1995) [13], has used genetic algorithms to evolve simulated neuro-organisms within a two-dimensional ecosystem. These experiments revealed that pre-adapted circuits can become more complex via the evolutionary process of reaching up. Reaching down, however, failed as an evolutionary strategy. Further experimental evidence supports this finding. Manipulations were designed to rule out alternative explanations, including: disruption of sensori-motor pathways during reaching down; the advantage of "mere" replication by reaching up; and, inability of network evolution to escape local mimima by reaching down. A reaching up complexification mutation was more likely to be successful if the new set of global connections had weights whose magnitudes were below average. Further experimental tests are proposed for the hypothesis that reaching up is an easier route to brain complexification to reaching down. | |
| Full Text |
|
|
Multimedia Links (none) Reference Links (none) Citation Reference |
Get viewers for PS & PDF   |
