Updating softmax in rethinking_2 chapter 11

[DamonCharlesRoberts]

Addresses #239

• Switches to pm.math.softmax() implementation for multinomial implementation of categorical model.
• Adds axis=1 to the function.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[ricardoV94]

Can we also get rid of the at imports. those were for aesara, now if anything it should be import pytensor.tensor as pt

[ricardoV94]

Also axis=-1, may be more readable (if it's indeed equivalent to axis=1)

[DamonCharlesRoberts]

Can we also get rid of the at imports. those were for aesara, now if anything it should be import pytensor.tensor as pt

Removed the at imports. Using the pymc.math.softmax instead for the application of the softmax function on objects within the model. For simulated data, left the original scipy import for the softmax function.

[DamonCharlesRoberts]

Also axis=-1, may be more readable (if it's indeed equivalent to axis=1)

Left as axis=1 to keep consistent style with other parts of the existing notebook.