Hi, is there a way to define a prior over initial values of an SDE used in Euler-Mayurama distribution?
For example I have defined the following SDE
def ou_sde(x, theta, alpha, s): "Ornstein-Uhlenbeck process" return alpha - theta * x, s
and passed it to the Euler-Mayurama distribution from timeseries models.
However, I can not figure out how to pass initial conditions for the integration process.
with pm.Model() as model: x0 = pm.Normal('x', 0., 1.) x = EulerMayurama('x', ou_sde, *args, init=x0)
The reason I am asking this is that by default both prior and posterior initial values make no sense, as I and up with some very large values for x0, far away from the mean of the process.