GitLab

Created by: J3FALL

The task is to implement functions for extended fedot experiments importing and exporting. A typical fedot use-case sounds like the following:

• You (data-scientist) have defined a ML-task/hypothesis: "To build a ML-pipeline for classification/regression/etc. with Fedot based on my own dataset"
• Prepare data according to fedot InputData format
• Configure Composer and run the optimization
• Obtain a suitable chain (pipeline) or set of chains
• Export the chain in json format

To improve the overall reproducibility of the experiments it would be useful to implement an utility function/class for exporting not only the resulted pipelines but also a dataset and logs (optional). And the same case but for importing. For instance, new_export-ing function can create a zip-archive/directory where dataset, chains and logs will be placed. Importing function should allow users to (a) obtain input data and chain from zip-file, (b) run fit/predict and show some CV-metrics.