Blind Simulation

Abstract

Consider the following problem. One is allowed to observe the output of a very complicated system. The system could be a computer model whose internal workings would contain many random number generators and various interacting subsystems. Alternatively one could have a physical device consisting of complicated possible nonlinear electronics from which one can sample data. It may even just be a string of numbers taken as data from some experiment. The simulation problem is to estimate the probability of some output sequence event. Because of the system’s complexity, it is not known nor would it be feasible analytically to derive the underlying probability law of the experiment. This is what we call a blind simulation problem: we have to estimate the probability of an event without complete knowledge of the underlying probability law. A very large class of practical simulation problems is blind.

God not only plays dice. He also sometimes throws them where they can’t be seen.

Stephen Hawking

Even a blind pig finds an acorn once in a while.

Anon