Wrapping a Model Optimizer library for WP9 in AIDApy

For the WP9 and its preliminary report which is due to 30/11 (Deliverable 9.1) our task is to add a model optimization library to aidapy. For this task we will initially try to integrate Nervana Distiller. The Interface of the Model Optimization module must be simple and Generic since in the future we will probably integrate and interconnect 2 other libraries for doing automatic hyperparameter optimization and CNN architecture search.

The tasks are:

1. Download the Time-series for a space mission.
2. Automatically split the xarray to train-val (and afterwards in a format that understands Nervana Distiller). @jannisteunissen may have you developed something with such functionality and if not are you planning to do it for the ml module ?
3. Make a first Model pruning or a quantization
4. Based on the above experience create an easy (and generic) interface
5. Try to add a new library