This is problematic in multinomial/softmax regression scenario’s though, for which you definitely would need p to support 0.
I can’t immediately think of a scenario with Dirichlet… but by analogy I would presume the same holds?
This is problematic in multinomial/softmax regression scenario’s though, for which you definitely would need p to support 0.
I can’t immediately think of a scenario with Dirichlet… but by analogy I would presume the same holds?