Federated learning improves privacy and accuracy in averaged dependence models
Federated Learning of AnDE Classifiers
Machine Learning
Summary
When computers learn from data gathered on different devices without sharing private details, this is called federated learning. The authors show how a special kind of probabilistic model called Averaged n-Dependence Estimators (AnDE) can be trained in such a way. Their method keeps private information safer by only sharing abstract information, not detailed data. Tests showed that their approach often works better than simpler classical methods, even when privacy protections are added.
What this means in practice
- •For mobile app developers: Train privacy-preserving models on users’ devices to improve app predictions without exposing private data.
- •For healthcare data teams: Aggregate patient data models across hospitals while protecting sensitive information and improving diagnosis predictions.
Authors
Pablo Torrijos, Juan C. Alfaro, José A. Gámez, José M. Puerta
Abstract
This work presents a federated framework for training Averaged $n$-Dependence Estimators (AnDE) in distributed environments. The proposed method focuses on the discriminative setting, where model weights are learned locally and aggregated globally, supporting any dependency order $n$. This design allows federated training without transmitting semantically meaningful parameters, improving privacy. Additionally, generative AnDE models are federated to provide a comparative baseline, with optional differential privacy applied to the aggregation of probability tables. Experiments on 12 discrete datasets show that discriminative models with $n \geq 1$ consistently outperform federated Naive Bayes (NB, $n=0$), and that privacy-preserving aggregation is effective with limited accuracy loss. These results establish federated AnDE as a viable and privacy-preserving framework, showing that probabilistic models remain applicable in modern federated learning settings.