didclab / RL-Optimizer

The RL optimization work by Jamil, Elvis, and Jacob in DIDCLAB
0 stars 2 forks source link

Use offline data for normalization #6

Closed elrodrigues closed 11 months ago

elrodrigues commented 1 year ago

Goal is to use offline 'gold' data to normalize observations when training. This is usually benchmarking data without any optimizers.

I have the data, so it's a matter of adding helpers to read, learn the statistics of the data, and pass those statistics to environment.

Note: see NetAcro's notebooks for hints on how to load and pick numeric columns.