Information in Mean Field Control

Thursday, November 12, 2020 - 1:40pm - 2:25pm
Aaron Palmer (University of British Columbia)
The dependence on available information is known to vanish for many `mean field' control problems. We investigate the role that information plays in the fluctuations about these mean field limits. In particular, we show how the fluctuations can be calculated efficiently for discrete mean field control problems with partial information, even when the number of states is large. We finish with a discussion of counterexamples that show the role of information does not vanish when the control problems have `short range' interactions.