How to make sure the distributions have same support?

numpyro
1 
import jax.numpy as jnp
from numpyro import distributions as dist

dist1 = dist.LogNormal(0, 1)

dist2 = dist.TransformedDistribution(
        dist.Chi2(2),
        dist.transforms.AffineTransform(0.0, 10, domain=dist1.support),
        )
print(dist1.support, dist2.support)

mixing = dist.Categorical(probs=jnp.array([0.2, 0.8]))
dist.Mixture(mixing, [dist1, dist2])

When running the above code, I got

ValueError: All component distributions must have the same support.

The support of the two distribution are Positive(lower_bound=0.0) and GreaterThan(lower_bound=0.0). How should I make they have same support?

2

you can use MixtureGeneral to provide a common support.

3

Thanks a lot, problem solved :tada: