Abstract:Federated Learning allows multiple participants to jointly train a global model while guaranteeing the confidentiality and integrity of private datasets. However, current server aggregation algorithms for federated learning only focus on model parameters, resulting in heavy communication costs and low convergence speed. Most importantly, they are unable to handle the scenario wherein different clients hold different local models with various network architectures. In this paper, we view these challenges from an alternative perspective: we draw attention to what should be aggregated and how to improve convergence efficiency. Specifically, we propose MHAT, a novel model-heterogenous aggregation training federated learning scheme which exploits a technique of Knowledge Distillation (KD) to extract the update information of the heterogenous model of all clients and trains an auxiliary model on the server to realize information aggregation. MHAT relaxes clients from fixing on an unified model architecture and significantly reduces the required computing resources while maintaining acceptable model convergence accuracy. Various experiments verify the effectiveness and applicability of our proposed scheme. (c) 2021 Elsevier Inc. All rights reserved. Federated learning 11?3] is an emerging machine learning paradigm for decentralized data 14,5], which enables multiple parties to collaboratively train a global model without sharing their private data. In the canonical federated learning protocol 16], model parameter is the interactive information between clients and the server. At the beginning of each training round, the central server distributes current global model to all participants, and all participants update the local model using their private data. Then the server will collect updated model information from all parties and obtain the central global model via a weighted average aggregation algorithm. However, in a practical federated learning, several major challenges still remain unsolved when using model parameters

Optimizing Multimodal Federated Learning: Novel Approaches for Efficient Model Aggregation and Client Sampling

Communication-Efficient Multimodal Federated Learning: Joint Modality and Client Selection

Multimodal Federated Learning

Multimodal federated learning: Concept, methods, applications and future directions

Multimodal Federated Learning on IoT Data

Federated Learning Model Training Method Based on Data Features Perception Aggregation

Resource-Efficient Federated Multimodal Learning via Layer-wise and Progressive Training

Towards Resource-Efficient and Secure Federated Multimedia Recommendation

Multi-Model Federated Learning

Adaptive Control of Client Selection and Gradient Compression for Efficient Federated Learning

Improving Global Generalization and Local Personalization for Federated Learning

Exploring System-Heterogeneous Federated Learning with Dynamic Model Selection

An Efficient Multi-Model Training Algorithm for Federated Learning

Unimodal Training-Multimodal Prediction: Cross-modal Federated Learning with Hierarchical Aggregation

Towards Efficient Federated Learning Framework Via Selective Aggregation of Models

A unified framework for multi-modal federated learning

Multimodal Fusion with Block Term Decomposition for Asynchronous Federated Learning

MHAT: an Efficient Model-Heterogenous Aggregation Training Scheme for Federated Learning.

Multimodal Federated Learning: A Survey

Communication Efficient Federated Learning With Heterogeneous Structured Client Models