GSO-2010-Shuai

GSO-2010: Xin Shuai and Larry Yaeger

Motif Analysis in Evolving Neural Networks

Networks motifs are called 'building block' in complex network. Motif analysis can give us richer and multi-dimensional information about the relations among complexity, structure and function. In this presentation, both the temporal dynamic and frequency distribution of 3-node motifs in Polyworld is analyzed for both driven runs and passive runs. Two interesting observations can be made based on the results: First, all 13 types of 3-node motifs can be classified into 6 categories according to their temporal dynamic trajectories. We hypothesize that some types of motifs are favored, while others are suppressed, by natural selection. Second, the top 3 frequent motif types are symmetric while the top 3 infrequency motif types are asymmetric. We hypothesize that symmetry in motif structure possibly has some evolutionary advantage in function. Further study is needed to validate our hypothesis.