2000-07 Week 1 (Suicide)


Early runs continued

stats1185: The fitness in these runs is again too high (around 4.6). I'm not sure what the story is (was). The simulation looks pretty similar to stats1157. I had added 2 more sets of statistics into these runs. One was health and age distributions for the entire agent population (not just those that suicide). These distributions were included for comparison to the same distributions amongst suiciding agents. I also added the distribution of genetic probabilities of suicide. These are averages over all genetic probabilities for a given act --- but regardless of rule, so I don't quite know what I was trying to do by adding this distribution to the statistics. Ah, I looked over my written notes, and I think these were added at the request of Kevin so that I could calculate the KL distance between the expected distribution of suicide probabilities under mutation, and the actual suicide probabilities that my simulation produced.