Most of the people who start out new don't find a latest feed of community hyped resources on ML and DL topics. It would be pretty good if we add a Reddit section.
If you're fine with this suggestion I'll put up a PR with the update 😄

While some methods do validate input data, some take data as-is and can result in runtime errors or unexpected behavior.

Formalizing the restrictions, both general dataset restrictions and valid meter topologies, will allow the addition of an optional validation step for each method.

(Potentially) related issues and comments:

• nilmtk/nilmtk#564 (comment)

The variable output_timeout is now a hardcoded constant:

https://github.com/computationalmodelling/nbval/blob/232d2496a01a5a5c96fe98b27977ae5505a15f5e/nbval/plugin.py#L336

For testing large outputs, such as notebook widgets, the 5 seconds are not sufficient.
Making the variable configurable would solve that issue.

In Search-As-You-Type, Completion, and Suggestion (Elasticsearch).ipynb, in the "Post-Search Suggest" section, suggest_body contains a small typo:

suggest_body = { 
    "title_comp

By either printing it off above results or using a global var that is set after each query like:

%%sql
select foo from bar

in another cell print _CE_ (or something):

> print(_CE_)
<engine-string> 

Should any other details be printed/returned? Should the object be returned to perform actions on it?