Simple Vertically Partitioned Model

[TTitcombe]

Description

Implement a simple SplitNN using syft which trains on vertically split data provided by multiple data holders

Why?

This may be the first open source implementation of a split model trained on vertically partitioned data. Training the model successfully is a complex process, which requires a custom data pipeline and model architecture.

This epic is the first stage of the project: The data will be ordered (no need for PSI) and we will send data labels away from the data holders

Breakdown

Provide a bulleted or numbered list of how you might break this epic down into smaller issues.

• Implement function to vertically split dataset (2 data holders/ 1 split) Function to vertically partition data #4
• Implement vertically split data loader Vertically Partitioned Data Loader #5
• Develop vertically split SplitNN Simple Vertically partitioned SplitNN #6

Who else?

May require changes in PySyft