Deep Latent Variable Model based Vertical Federated Learning with Flexible Alignment and Labeling Scenarios

2025-05-16

AI Summary

Existing vertical federated learning (VFL) methods are constrained by strict assumptions on participant count, data alignment quality, and label availability, limiting their applicability to real-world scenarios featuring diverse missing-data patterns and heterogeneous alignment structures. To address this, we propose the first unified VFL framework supporting arbitrary data alignment configurations, arbitrary label missingness mechanisms, and multi-party collaboration. Our approach formalizes alignment gaps as missing-data problems and leverages deep latent-variable models—specifically VAEs and VAEGANs—to jointly perform generative modeling, probabilistic inference, and privacy-preserving federated optimization. The framework accommodates unlabeled samples, dynamically adapts to varying alignment degrees, and enables multi-stage collaborative training. Evaluated across 168 configurations, it outperforms all baselines in 160 cases, achieving an average accuracy gain of 9.6 percentage points. Empirical results on healthcare and financial heterogeneous benchmarks demonstrate substantial improvements in practical utility and generalization capability.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Multiagent Systems: Adversarial Agents

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Federated learning (FL) has attracted significant attention for enabling collaborative learning without exposing private data. Among the primary variants of FL, vertical federated learning (VFL) addresses feature-partitioned data held by multiple institutions, each holding complementary information for the same set of users. However, existing VFL methods often impose restrictive assumptions such as a small number of participating parties, fully aligned data, or only using labeled data. In this work, we reinterpret alignment gaps in VFL as missing data problems and propose a unified framework that accommodates both training and inference under arbitrary alignment and labeling scenarios, while supporting diverse missingness mechanisms. In the experiments on 168 configurations spanning four benchmark datasets, six training-time missingness patterns, and seven testing-time missingness patterns, our method outperforms all baselines in 160 cases with an average gap of 9.6 percentage points over the next-best competitors. To the best of our knowledge, this is the first VFL framework to jointly handle arbitrary data alignment, unlabeled data, and multi-party collaboration all at once.

Problem

• Handles feature-partitioned data with alignment gaps
• Supports unlabeled data and multi-party collaboration
• Addresses diverse missingness mechanisms in VFL

Innovation

• Deep Latent Variable Model for VFL
• Handles arbitrary alignment and labeling
• Supports multi-party missing data