Communication is widely known as the primary bottleneck of federated learning, and quantization of local model updates before uploading to the parameter server is an effective solution to reduce the communication overhead. However, prior literature always assumes homogeneous quantization for all clients, while in reality, devices are heterogeneous and support different levels of quantization precision. This heterogeneity of quantization poses a new challenge: fine-quantized model updates are more accurate than coarse-quantized ones, and how to optimally aggregate them at the server is an open problem. In this paper, we propose FedHQ – Federated Learning with Heterogeneous Quantization – that allocates different aggregation weights to different clients by minimizing the convergence rate upper bound as a function of the heterogeneous quantization errors of all clients, for both strongly convex and non-convex loss functions. To further accelerate the convergence, the instantaneous quantization error is computed and piggybacked when each client uploads the local model update, and the server dynamically calculates the weight accordingly for the current aggregation. Numerical experiment results demonstrate the performance advantages of FedHQ over both vanilla FedAvg with standard equal weights and a heuristic aggregation scheme, which assigns weights linearly proportional to the clients’ quantization precision.

Vehicular ad hoc networks (VANETs) is one of the most promising approaches for the Intelligent Transportation Systems (ITS). With the rapid increase in the amount of traffic data, deep learning based algorithms have been used extensively in VANETs. The recently proposed federated learning is an attractive candidate for collaborative machine learning where instead of transferring a plethora of data to a centralized server, all clients train their respective local models and upload them to the server for model aggregation. Model quantization is an effective approach to address the communication efficiency issue in federated learning, and yet existing studies largely assume homogeneous quantization for all clients. However, in reality, clients are predominantly heterogeneous, where they support different quantization precision levels. In this work, we propose FedDO – Federated Learning with Double Optimization. Minimizing the drift term in the convergence analysis, which is a weighted sum of squared quantization errors (SQE) over all clients, leads to a double optimization at both clients and server sides. In particular, each client adopts a fully distributed, instantaneous (per learning round) and individualized (per client) quantization scheme that minimizes its own squared quantization error, and the server computes the aggregation weights that minimize the weighted sum of squared quantization errors over all clients. We show via numerical experiments that the minimal-SQE quantizer has a better performance than a widely adopted linear quantizer for federated learning. We also demonstrate the performance advantages of FedDO over the vanilla FedAvg with standard equal weights and linear quantization.

In recent years, Graph Neural Networks (GNNs) based on deep learning techniques have achieved promising results in EEG-based depression detection tasks but still have some limitations. Firstly, most existing GNN-based methods use pre-computed graph adjacency matrices, which ignore the differences in brain networks between individuals. Additionally, methods based on graph-structured data do not consider the temporal dependency information of brain networks. To address these issues, we propose a deep learning algorithm that explores adaptive graph topologies and temporal graph networks for EEG-based depression detection. Specifically, we designed an Adaptive Graph Topology Generation (AGTG) module that can adaptively model the real-time connectivity of the brain networks, revealing differences between individuals. In addition, we designed a Graph Convolutional Gated Recurrent Unit (GCGRU) module to capture the temporal dynamical changes of brain networks. To further explore the differential features between depressed and healthy individuals, we adopt Graph Topology-based Max-Pooling (GTMP) module to extract graph representation vectors accurately. We conduct a comparative analysis with several advanced algorithms on both public and our own datasets. The results reveal that our final model achieves the highest Area Under the Receiver Operating Characteristic Curve (AUROC) on both datasets, with values of 83% and 99%, respectively. Furthermore, we perform extensive validation experiments demonstrating our proposed method’s effectiveness and advantages. Finally, we present a comprehensive discussion on the differences in brain networks between healthy and depressed individuals based on the outputs of our final model’s AGTG and GTMP modules.

Nowadays, Federated Learning (FL) over Internet of Medical Things (IoMT) devices has become a current research hotspot. As a new architecture, FL can well protect the data privacy of IoMT devices, but the security of neural network model transmission can not be guaranteed. On the other hand, the sizes of current popular neural network models are usually relatively extensive, and how to deploy them on the IoMT devices has become a challenge. One promising approach to these problems is to reduce the network scale by quantizing the parameters of the neural networks, which can greatly improve the security of data transmission and reduce the transmission cost. In the previous literature, the fixed-point quantizer with stochastic rounding has been shown to have better performance than other quantization methods. However, how to design such quantizer to achieve the minimum square quantization error is still unknown. In addition, how to apply this quantizer in the FL framework also needs investigation. To address these questions, in this paper, we propose FEDMSQE – Federated Learning with Minimum Square Quantization Error, that achieves the smallest quantization error for each individual client in the FL setting. Through numerical experiments in both single-node and FL scenarios, we prove that our proposed algorithm can achieve higher accuracy and lower quantization error than other quantization methods.