We mentioned earlier that our choice of the probability distribution doesn't impact Monte Carlo Integration, but the choice of p(w) here seems pretty deliberate. Would we have been able to accomplish the same thing with some other distribution on w?