New Operator

Describe the operator

Why is this operator necessary? What does it accomplish?

This is a frequently used operator in tensorflow/keras

Can this operator be constructed using existing onnx operators?

If so, why not add it as a function?

I don't know.

Is this operator used by any model currently? Which one?

Are you willing to contribute it?

I noticed our release version anchor links in the changelog don't actually reference a specific released version. If I go to the changelog and click on the 2021.12.0 link, I'm redirected to https://docs.dask.org/en/stable/changelog.html#id1 when, naively, I would have expected this link to look like https://docs.dask.org/en/stable/changelog.html#2021.12.0 (or something similar). As you move down

The components part of our codebase was written sometime ago, with older sklearn versions and before python typing was production ready.

In general, some of these files need to be cleaned up. Mostly typing of parameters and functions, adding documentation a bout these parameters and finally double checking with scikit learn that there aren't some new or deprecated parameters we still use.

To

• With Featuretools 1.0.0 we add a dataframe to an EntitySet with the following:

es = ft.EntitySet('new_es')

es.add_dataframe(dataframe=orders_df,
                 dataframe_name='orders',
                 index='order_id',
                 time_index='order_date')

Improvement

• However, you could also change the EntitySet setter to add it with this approach:

es = ft.Ent

The extension templates (here: https://github.com/alan-turing-institute/sktime/tree/main/extension_templates) should be extended with a preamble that treats soft dependencies.

The challenge is to keep it very brief, and to make clear that adding soft dependencies is only necessary for the case of extending sktime itself. If the template is used in a project/package that has sktime as a depe

Yes

The current History class has some limitations: (ver 0.10.0)

1. Currently the history is saved as JSON, as a result, those recorded values are limited to simple numbers and strings. Other objects can not be saved in history files directly.
2. Saving as JSON takes lots of time and space because numbers are stored in decimal. It's getting worse when the training epoch is increasing.
3. In some

Add documentation (initially just for the TabularPredictor) explaining what feature engineering AutoGluon does for the user by default:

• what is the universe of types each column can be
• how do we infer types per column
• for each type, what further feature engineering is done (e.g. ngrams and the date parsing I just submitted)
• what are the parameters used for the heuristics
• how does

Can we have an example of REST API calls in the documentation?
Examples with CURL, HTTPie or another client and the results would be better for newbies.
Thanks again for your good work.