Previous | Next --- Slide 12 of 52
Back to Lecture Thumbnails

How do you know the exact probability distribution for something unknown?


What are certain scenarios where we wouldn't want to bias the samples, like maybe if we have enough lighting everywhere or maybe could the machine have enough compute to make taking less beneficial samples not that wasteful overall?


In what situations would we sample non-uniformly and how would we determine our probability distribution?


What would be the purpose of sampling non-uniformly if we wanted to weigh them into be uniform anyways?


So would you want to sample non-uniformly when it costs less to do so?


Are these probability distributions estimated? When do we want to sample non-uniformly vs uniformly?