Ricardo Vieira
Ricardo Vieira
### Description As shown in one of the new examples in #7014 ```python import pymc as pm with pm.Model() as model: x = pm.Normal("x") y = pm.Normal("y") det_xy = pm.Deterministic("det_xy",...
### Description sample calls it draws, and the default dim is also `draw`. Whenever I want to use a kwarg to specify the number of draws I am always surprised...
We should release again with the patch from #61 It looks like the last [conda](https://github.com/conda-forge/nutpie-feedstock/pull/22) version was released without updating the python version, so it shows as 0.8.1 while it...
Would be neat if we could use nutpie as a step sampler with the other PyMC samplers (without killing performance too much) Might justify refactoring the trace backend CC @michaelosthege...
Reported by @jessegrabowski ```python with MarginalModel(coords=coords) as m: x_data = pm.ConstantData('x', df.x, dims=['obs_idx']) y_data = pm.ConstantData('y', df.y, dims=['obs_idx']) X = pt.concatenate([pt.ones_like(x_data[:, None]), x_data[:, None], x_data[:, None] ** 2], axis=-1) mu...
A subclass of the Categorical sampler from PyMC should suffice
```python import pymc as pm from pymc_experimental import MarginalModel with MarginalModel() as m: x = pm.Bernoulli("x", p=0.5) y = pm.Normal("y", x) m.marginalize(["x"]) m.unmarginalize(["x"]) # ValueError: list.remove(x): x not in list...
This PR allows more kinds of graphs to be marginalized. Previously, we were limiting it to Elemwise operations to ensure that information across batch dimensions was not mixed between the...
> Not a blocker, but we should be able to support this. Here's a paper with equations for a 2nd order HMM, we should be able to work out the...
#91 Implements automatic logp marginalization for bernoulli, categorical and discrete uniform, as those are the only default discrete univariate RVs with finite support. We could extend it easily to right-truncated...