Creating a Federated Learning experiment

When you and your remote parties complete the previous set up, you can proceed to launch the Federated Learning experiment and configure its settings for the run.

Cloud open beta

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

Start Federated Learning

  1. In your project home, click Add to Project > Federated Learning.
  2. In the Define details tab, fill out the required fields including the training Name, an optional Description, and you can also add tags if so desired.
  3. In the Configure tab, choose the training framework and its model type. See Choosing your framework and fusion method for a table of frameworks, fusion methods and their details. Click Next.
  4. Click the Define hyperparameters tab to specify how to tune your model, the hyperparameters available are based on your framework and fusion method. See Hyperparameters configuration for more details.
  5. In the Select remote training system tab, select any Remote Training System that you previously created. You can filter Remote Training Systems with tags so that you can see specific ones that you want. If you have not created a Remote Training System, follow this step to create one: Creating a Remote Training System.
  6. Click on Review and create to view the settings of your current Federated Learning experiment. Then click on Create.

Next steps

In the preparation stage of the Federated Learning experiment’s running UI, click View setup information to view details of this experiment. Proceed to Preparing the parties’ configuration.