Normalize subgroups about transform HOT 3 CLOSED

We are working on group-by-key based analyzers. Do you have a specific use case in mind? tf-idf is not implemented with a group-by-key approach, so you could potentially copy its implementation for your own use case. Generally we would expected group-by-key based analyzers to be faster but they may not have any advantage when the data being aggregated is small, as would typically be the case for a numeric analyzer.

from transform.

Yes, the use case is to normalize energy use per customer. If the most important metric is absolute error, then this is not necessary, but if it's about predicting the shape of every load curve as accurately as possible this might be a good transformation. We need this mainly to experiment though, and have switched to using Dask for computing statistics and doing preprocessing as it is much more flexible. We're still stealing from TFT the idea of embedding these normalization stats in the TF graph, specifically we're using the normalizer_fn kwarg of tf.feature_column.numeric_column.

from transform.

This question is better asked on StackOverflow since it is not a bug or feature request. There is also a larger community that reads questions there.

If you think we've misinterpreted a bug, please comment again with a clear explanation, as well as all of the information requested in the issue template. Thanks!

from transform.