Most functions in scipy.linalg functions (e.g. svd, qr, eig, eigh, pinv, pinv2 ...) have a default kwarg check_finite=True that we typically leave to the default value in scikit-learn.

As we already validate the input data for most estimators in scikit-learn, this check is redundant and can cause significant overhead, especially at predict / transform time. We should probably a

Collection of follow-ups to #5827. These can/should be broken out into individual PRs. Many are relatively straightforward and would make a good first PR.

General

• Documentation (none was added in original PR).
• Release notes.
• Example notebook.
• Double-check how sm.tsa.arima.ARIMA works with fix_params (it should fail except when the fit method is statespace

Improve examples such that they are more incremental (in the import etc) without following strictly PEP8. It will make it nicer to read on the gallery generated online.

See:
gonum/gonum#845 (comment)

In X-ray crystallography, the most important prior distributions include two special cases of the generalized gamma distribtion. I am very keen to try this parameterization of the variational distritribution in my research project. How hard would it be for the TFP devs to implement this distr

1. "Conclusion" section of "Getting started with Tablesaw" page contains broken link to "Java Docs".
   https://jtablesaw.github.io/tablesaw/gettingstarted#conclusion
2. "Exploring tables" section of "Getting started with Tablesaw" page contains broken link to "plotting".
   https://jtablesaw.github.io/tablesaw/gettingstarted.html#exploring-tables

I found several useful applications of pseudo-random number sampling in the past. In particular:

• Inverse transform sampling
• Gibbs sampling

(This issue serves a reminder to add the respective methods. Pull requests always welcome.)

Since the default output is meant to be human-readable, would it make sense to add thousands separators to make the output more easily readable?