Workflow-Ant Privacy-Preserving Computation Platform-阿里云帮助中心

This topic describes the overall workflow for federated model development and explains how to develop federated models in the federated modeling console.

Federated modeling is a component of the Ant Privacy-Preserving Computation Platform. Before you can use the federated modeling console, you must first deploy federated learning nodes. Then, you can register and configure the nodes and projects in the Management Console. The following graph illustrates this process.

The federated modeling workflow is as follows:

Deploy federated learning nodes and load sample data into the nodes. For more information, see Install and deploy nodes.
In the Management Console, perform the following steps:
1. Register the federated learning nodes. For more information, see Create a node.
2. Register the sample data. For more information, see Data tables.
3. Create a federated learning (FL) project. For more information, see Create a project.
4. Grant the federated nodes access to the FL project. For more information, see Step 1: Configure nodes.
5. Grant the data on the federated nodes access to the FL project. For more information, see Step 2: Configure data.
6. Add members to the FL project. For more information, see Step 3: Configure members.
The federated model development workflow in the Federated Modeling Console is as follows:
For steps 1 to 5, see Create a script.
1. Create a federated table
2. Data pre-processing
3. Model training
4. Preprocessing application
5. Model evaluation
6. Save the model. For more information, see Save a model.

Note

For more information about federated model development, see Federated modeling.