Vibrational control of chaos in artificial neural networks

摘要: Neural networks with chaotic baseline behavior are interesting for their experimental bases in both biological relevancy and engineering applicability. In the case, literature still lacks a robust study of interrelationship between particular network dynamics “online” or “driving” inputs. We ask question, neural behaviour, what periodic inputs minimal magnitude have stabilizing effect on dynamics? A genetic algorithm is developed task. systematic comparison different operators carried out where each operator-combination ranked by optimality solutions found. The reaches acceptable results finds input sequences largest elements order 10−3. Lastly, an illustration complexity fitness space produced brute-force sampling period-2 plotting map network. Thesis Committee Adviser: Dr. Roger Gaborski Observer: Peter Anderson Reader: Thomas Borelli