Summary

• Convergence/Consensus if and only if aperiodicity

• Limiting influence related to eigenvectors and weights from influential neighbors

• Wise crowds: nobody retains too much influence

Learning Models

• Bayesian is computationally demanding in network settings

• Restricted Bayesian gives consensus network not much of a role

• DeGroot and other myopic models bring network into play

• Can reach consensus, can be wise

• Influence and speed are tractable...