Using pytensor random number generator with scan for Simulator class

Questions v5
8

Is it unfeasible to marginalize the DiscreteMarkovChain? That’s what we would recommend (you never actually want to sample it). pm.marginalize works with chains with lags=1