Ricardo Vieira

Implement multi-output Elemwise in Numba via guvectorize

2

This was just a quick hack to see if it makes sense. I have not worked with Numba before and I mostly pattern-matched my way so far. I made it...

Scaled Dirichlet does not integrate to 1

1

I am not sure if it should, but when I scale a Dirichlet non-homogeneously, it seems to no longer integrate to 1: ```python import numpy as np import tensorflow_probability as...

BUG: Behavior of `atleast_3d` is surprising

7

### Describe the issue: Adding axis to the right is perplexing, specially when `atleast_2d` adds to the left and numpy broadcasting behavior always grows to the left. Is there a...

ENH:linalg: kron does not support nd arrays with n>2

11

Is this a feature? > Re. `scipy.linalg.kron`, I'm not sure whether it might not be better to just delegate to `numpy.kron`, unless there are any advantages to `scipy.linalg.kron`. `numpy.kron` works...

Update blackbox likelihood example

22

## Changes - Make Op more generic (accepts distinct inputs, returns vectorized loglikelihood), for better transfer learning - Build Ops more iteratively, and show useful debugging strategies - Use CustomDist...

Docs link: https://www.pymc.io/projects/examples/en/latest/variational_inference/pathfinder.html The notebook is not using the parameters `mu` or `tau`. It creates a `theta1` that is not used anywhere. However, I am not sure pathfinder works with...

`DensityDist` is now just a backwards-compatible alias and is not even in the documentation (although maybe it should?)

The latest release of PyMC allows for `Model(mutable_coords=...)` which makes Out of sample prediction with "coordinated" variables more straightforward. We should update any relevant examples. This is a bit of...

Make use of numba opt-in not opt-out

3

Calling `forest_plot` when numba is installed ends up with cryptic errors: `ValueError: Failed in object mode pipeline (step: remove phis nodes)` This is a PITA. Please consider making the use...

How to define the location of plot_forest legend?

3

It tends to overlap with the posterior lines. I would like to set up something like `loc=(1, 0.5)`