I come from an R background. I want to implement smooth terms in pyro.
In R, I would code something like this:
m <- lm(y ~ x1 + bs(x2, degree = 3, knots = c(0, 10))
which generally works the same way in Stan/brms:
m <- model <- brm(y ~ x1 + bs(x2, degree = 3, knots = c(0, 10), family = gaussian(), prior = prior, algorithm = "fullrank" )
I’m trying to do this in pyro. Is there any corresponding functionality to make smooth terms?
This is the only post about splines I’ve seen on pyro. This seems to suggest that implementing the model with a spline in sklearn, extracting the design matrix, and then using the sklearn parameter estimates as my priors is probably my best (only?) path forward. Is this accurate? Or is there a more straightforward method in pyro?
Thanks in advance!