HFL Model Training Overview

Horizontal Federated Learning (HFL) is used in scenarios where datasets share the same feature space (same type of columns) but differ in samples (different rows).

Key features of HFL:

• Data never leaves client nodes.

• Datasets have the same features, but different records. All the clients hold label information.

• Computation happens client-side; server only does aggregation.

• After training, each client can get a copy of the global model.

HFL example use case

An example of an HFL use case is Insurance Home Inspection.

A insurance carrier has a model predicting the number of issues that will be identified through home inspection in specific regions in Canada. They would like to expand their model’s scope by using a data provider’s nationwide inspection data sets and determine model boost before purchasing.

Available HFL models

integrate.ai supports the following HFL model types:

• FFNet - Feed forward neural network

• Generalized linear models (GLM)

• Gradient Boosted Models (GBM)

• Making predictions

Vertical Federated Learning (VFL) is also supported. Click here to learn more.