Skip to main content
Advertisement

< Back to Article

Implications of Behavioral Architecture for the Evolution of Self-Organized Division of Labor

Figure 6

Recurrent neural networks: Evolutionary trajectories of network parameters leading to experience-based specialization.

A simulation is shown in which all colonies in the population evolve high degree of division of labor. Parameter values are: ½, , . The self-feedback connection weights, and (third and fourth graph on the top panel), increase over generations, a pattern which is found across simulations showing the same worker specialization patterns. Also representative is the pattern encountered in the other connection weights (bottom panel) is the evolution of negative values in cross-connection weights ( and ) and positive values in the connection weights between the task stimulus and respective output node ( and ).

Figure 6

doi: https://doi.org/10.1371/journal.pcbi.1002430.g006