Sampling without replacement with unequal, dependent probabilities
问题 So from answers to the question, Randomly Generating Combinations From Variable Weights, I've managed to sample without replacement given unequal probabilities. I've implemented this in my particular context and it works great. Now I have access to additional information about my distribution. Specifically, I've been using a neural network to generate these probabilities so far, and I've now trained my neural network to output a probability for each pair of unique objects, in addition to