basically, if I change the input by X percent, the output will change by Y percent

the conditioning ratio number is the ratio between Y and X it's obtained when picking infinitesimally small X

When the ray goes through the medium, it may do a surface interactions with it. Tr represents the probability that there will be no interactions. However, some of the interactions in delta tracking might be false (because they're interactions with virtual particles). Thus, the probabilty to reach the end of the medium is equal to the product of surviving each interaction until reaching the end of the medium. As such, the probability is the product of density(p) / maxDensity for all points p of interaction. This is much more efficient than executing sample() many times and then seeing on average how many times there has been no interaction with the medium.

Steps: 1) Use a uniform sample to determine the channel i we'll use 2) Pick t using the corresponding sigma(t, i) 3) Use that t to compute the average

This avoids using different t's for different channels