Abstract:E-health allows smart devices and medical institutions to collaboratively collect patients' data, which is trained by Artificial Intelligence (AI) technologies to help doctors make diagnosis. By allowing multiple devices to train models collaboratively, federated learning is a promising solution to address the communication and privacy issues in e-health. However, applying federated learning in e-health faces many challenges. First, medical data is both horizontally and vertically partitioned. Since single Horizontal Federated Learning (HFL) or Vertical Federated Learning (VFL) techniques cannot deal with both types of data partitioning, directly applying them may consume excessive communication cost due to transmitting a part of raw data when requiring high modeling accuracy. Second, a naive combination of HFL and VFL has limitations including low training efficiency, unsound convergence analysis, and lack of parameter tuning strategies. In this paper, we provide a thorough study on an effective integration of HFL and VFL, to achieve communication efficiency and overcome the above limitations when data is both horizontally and vertically partitioned. Specifically, we propose a hybrid federated learning framework with one intermediate result exchange and two aggregation phases. Based on this framework, we develop a Hybrid Stochastic Gradient Descent (HSGD) algorithm to train models. Then, we theoretically analyze the convergence upper bound of the proposed algorithm. Using the convergence results, we design adaptive strategies to adjust the training parameters and shrink the size of transmitted data. Experimental results validate that the proposed HSGD algorithm can achieve the desired accuracy while reducing communication cost, and they also verify the effectiveness of the adaptive strategies.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the low communication efficiency and privacy protection issues in e - health due to horizontal and vertical data partitioning. Specifically, the challenges mentioned in the paper include: 1. **Horizontal and vertical data partitioning**: Medical data has both horizontal partitioning (samples are different but features are the same among different devices or hospitals) and vertical partitioning (the data of the same patient is distributed among different devices and hospitals). Using horizontal federated learning (HFL) or vertical federated learning (VFL) alone cannot effectively handle these two types of partitioning, and direct application will lead to excessive communication costs. 2. **Limitations of existing methods**: - The method of simply combining HFL and VFL has deficiencies in training efficiency, convergence analysis, and parameter adjustment strategies. - The interaction between HFL and VFL complicates the theoretical convergence proof and lacks the convergence results required to optimize the training process. - The lag of intermediate results affects the modeling efficiency, especially in the global multi - classification method. - Training efficiency is affected by multiple configurations, such as global aggregation interval, local communication frequency, and learning rate, but the specific impact of these parameters on model performance is still unclear. To solve these problems, the author proposes a new hybrid federated learning framework, aiming to improve communication efficiency and overcome the above limitations. This framework includes an intermediate result exchange stage and two aggregation stages (local aggregation and global aggregation), and a hybrid stochastic gradient descent (HSGD) algorithm is developed to train the model. In addition, the author also conducts theoretical analysis, derives the convergence conditions of the algorithm, and designs an adaptive strategy to adjust training parameters, thereby reducing communication costs and improving accuracy.

Communication-Efficient Hybrid Federated Learning for E-health with Horizontal and Vertical Data Partitioning

A Federated Learning Framework Via Decentralized Data Valuation for Chronic Disease Healthcare

FedDGP: Disentangling Global and Personal Models for Federated Learning

Hierarchical Federated Learning with Adaptive Clustering on Non-IID Data

A Communication-efficient Federated Learning Assisted by Central Data: Implementation of Vertical Training into Horizontal Federated Learning

FedEmb: A Vertical and Hybrid Federated Learning Algorithm using Network And Feature Embedding Aggregation

Enhanced Hybrid Hierarchical Federated Edge Learning Over Heterogeneous Networks

A Joint Communication and Federated Learning Framework for Internet of Things Networks

Communication-efficient hierarchical federated learning for IoT heterogeneous systems with imbalanced data

A Communication-Efficient Hierarchical Federated Learning Framework via Shaping Data Distribution at Edge

A Communication Efficient Vertical Federated Learning Framework.

Communication-Efficient Personalized Federated Learning for Green Communications in IoMT

Importance-aware Data Selection and Resource Allocation for Hierarchical Federated Edge Learning

HSFL: an Efficient Split Federated Learning Framework Via Hierarchical Organization.

A Scalable Federated Learning Approach for Collaborative Smart Healthcare Systems with Intermittent Clients using Medical Imaging

Channel Adaptive and Sparsity Personalized Federated Learning for Privacy Protection in Smart Healthcare Systems

SHARE: Shaping Data Distribution at Edge for Communication-Efficient Hierarchical Federated Learning

Accelerating Hierarchical Federated Learning with Model Splitting in Edge Computing

Optimizing Training Efficiency and Cost of Hierarchical Federated Learning in Heterogeneous Mobile-Edge Cloud Computing

A Federated Learning Framework for Healthcare IoT devices

Unified Fair Federated Learning for Digital Healthcare