Getting set up for Federated Learning

Before you can get started with Federated Learning, you must plan out some of the components.

Cloud open beta

This is a Cloud open preview and is not supported for use in production environments.

Here is what you need to set up before proceeding with Federated Learning.

  1. Federated Learning automatically comes with the IBM Watson Machine Learning service. See: Setting up your environment for IBM Watson Machine Learning.
  2. Create a project. To learn more about projects, see Projects.
  3. Federated Learning is designed to connect data from multiple parties but at least one party must exist with:
    • A remote server where they run their local training that can be connected to a Remote Training System. To view the requirements for being able to connect to the Remote Training System, see Remote server requirements.
    • Data to be used for training on their remote training system. The data can be of a wide variety of formats, including databases, CSV files.
  4. Download the Federated Learning Python SDK, which includes these necessary files for training:
    • A party configuration yml file
    • A training handler file
  5. All parties involved must be collaborators to the Watson Studio project where the Federated Learning experiment exists.
  6. Note: For all parties, they must follow the steps documented in Anaconda configuration to set up their Anaconda environment.

Next step

Now you can get started with Federated Learning.