Adaptive Load Balancing for Parameter Servers in Distributed Machine Learning over Heterogeneous Networks

In distributed machine learning(DML)based on the parameter server(PS)architecture,unbalanced communication load distribution of PSs will lead to a significant slowdown of model synchronization in heterogeneous networks due to low utilization of bandwidth.To address this problem,a network-aware adaptive PS load distribution scheme is proposed,which accelerates model synchronization by proactively adjusting the communication load on PSs according to network states.We evaluate the proposed scheme on MXNet,known as a realworld distributed training platform,and results show that our scheme achieves up to 2.68 times speed-up of model training in the dynamic and heterogeneous network environment.

Pseudo-label based semi-supervised learning in the distributed machine learning framework

With the emergence of various intelligent applications,machine learning technologies face lots of challenges including large-scale models,application oriented real-time dataset and limited capabilities of nodes in practice.Therefore,distributed machine learning(DML) and semi-supervised learning methods which help solve these problems have been addressed in both academia and industry.In this paper,the semi-supervised learning method and the data parallelism DML framework are combined.The pseudo-label based local loss function for each distributed node is studied,and the stochastic gradient descent(SGD) based distributed parameter update principle is derived.A demo that implements the pseudo-label based semi-supervised learning in the DML framework is conducted,and the CIFAR-10 dataset for target classification is used to evaluate the performance.Experimental results confirm the convergence and the accuracy of the model using the pseudo-label based semi-supervised learning in the DML framework.Given the proportion of the pseudo-label dataset is 20%,the accuracy of the model is over 90% when the value of local parameter update steps between two global aggregations is less than 5.Besides,fixing the global aggregations interval to 3,the model converges with acceptable performance degradation when the proportion of the pseudo-label dataset varies from 20% to 80%.

A Tutorial on Federated Learning from Theory to Practice:Foundations,Software Frameworks,Exemplary Use Cases,and Selected Trends

When data privacy is imposed as a necessity,Federated learning(FL)emerges as a relevant artificial intelligence field for developing machine learning(ML)models in a distributed and decentralized environment.FL allows ML models to be trained on local devices without any need for centralized data transfer,thereby reducing both the exposure of sensitive data and the possibility of data interception by malicious third parties.This paradigm has gained momentum in the last few years,spurred by the plethora of real-world applications that have leveraged its ability to improve the efficiency of distributed learning and to accommodate numerous participants with their data sources.By virtue of FL,models can be learned from all such distributed data sources while preserving data privacy.The aim of this paper is to provide a practical tutorial on FL,including a short methodology and a systematic analysis of existing software frameworks.Furthermore,our tutorial provides exemplary cases of study from three complementary perspectives:i)Foundations of FL,describing the main components of FL,from key elements to FL categories;ii)Implementation guidelines and exemplary cases of study,by systematically examining the functionalities provided by existing software frameworks for FL deployment,devising a methodology to design a FL scenario,and providing exemplary cases of study with source code for different ML approaches;and iii)Trends,shortly reviewing a non-exhaustive list of research directions that are under active investigation in the current FL landscape.The ultimate purpose of this work is to establish itself as a referential work for researchers,developers,and data scientists willing to explore the capabilities of FL in practical applications.

Deep reinforcement learning based worker selection for distributed machine learning enhanced edge intelligence in internet of vehicles

Nowadays,Edge Information System(EIS)has received a lot of attentions.In EIS,Distributed Machine Learning(DML),which requires fewer computing resources,can implement many artificial intelligent applications efficiently.However,due to the dynamical network topology and the fluctuating transmission quality at the edge,work node selection affects the performance of DML a lot.In this paper,we focus on the Internet of Vehicles(IoV),one of the typical scenarios of EIS,and consider the DML-based High Definition(HD)mapping and intelligent driving decision model as the example.The worker selection problem is modeled as a Markov Decision Process(MDP),maximizing the DML model aggregate performance related to the timeliness of the local model,the transmission quality of model parameters uploading,and the effective sensing area of the worker.A Deep Reinforcement Learning(DRL)based solution is proposed,called the Worker Selection based on Policy Gradient(PG-WS)algorithm.The policy mapping from the system state to the worker selection action is represented by a deep neural network.The episodic simulations are built and the REINFORCE algorithm with baseline is used to train the policy network.Results show that the proposed PG-WS algorithm outperforms other comparation methods.

The adaptive distributed learning based on homomorphic encryption and blockchain

The privacy and security of data are recently research hotspots and challenges.For this issue,an adaptive scheme of distributed learning based on homomorphic encryption and blockchain is proposed.Specifically,in the form of homomorphic encryption,the computing party iteratively aggregates the learning models from distributed participants,so that the privacy of both the data and model is ensured.Moreover,the aggregations are recorded and verified by blockchain,which prevents attacks from malicious nodes and guarantees the reliability of learning.For these sophisticated privacy and security technologies,the computation cost and energy consumption in both the encrypted learning and consensus reaching are analyzed,based on which a joint optimization of computation resources allocation and adaptive aggregation to minimize loss function is established with the realistic solution followed.Finally,the simulations and analysis evaluate the performance of the proposed scheme.

FedTC:A Personalized Federated LearningMethod with Two Classifiers

Centralized training of deep learning models poses privacy risks that hinder their deployment.Federated learning(FL)has emerged as a solution to address these risks,allowing multiple clients to train deep learning models collaborativelywithout sharing rawdata.However,FL is vulnerable to the impact of heterogeneous distributed data,which weakens convergence stability and suboptimal performance of the trained model on local data.This is due to the discarding of the old local model at each round of training,which results in the loss of personalized information in the model critical for maintaining model accuracy and ensuring robustness.In this paper,we propose FedTC,a personalized federated learning method with two classifiers that can retain personalized information in the local model and improve the model’s performance on local data.FedTC divides the model into two parts,namely,the extractor and the classifier,where the classifier is the last layer of the model,and the extractor consists of other layers.The classifier in the local model is always retained to ensure that the personalized information is not lost.After receiving the global model,the local extractor is overwritten by the globalmodel’s extractor,and the classifier of the globalmodel serves as anadditional classifier of the localmodel toguide local training.The FedTCintroduces a two-classifier training strategy to coordinate the two classifiers for local model updates.Experimental results on Cifar10 and Cifar100 datasets demonstrate that FedTC performs better on heterogeneous data than current studies,such as FedAvg,FedPer,and local training,achieving a maximum improvement of 27.95%in model classification test accuracy compared to FedAvg.

Distributed Deep Reinforcement Learning:A Survey and a Multi-player Multi-agent Learning Toolbox

With the breakthrough of AlphaGo,deep reinforcement learning has become a recognized technique for solving sequential decision-making problems.Despite its reputation,data inefficiency caused by its trial and error learning mechanism makes deep reinforcement learning difficult to apply in a wide range of areas.Many methods have been developed for sample efficient deep reinforcement learning,such as environment modelling,experience transfer,and distributed modifications,among which distributed deep reinforcement learning has shown its potential in various applications,such as human-computer gaming and intelligent transportation.In this paper,we conclude the state of this exciting field,by comparing the classical distributed deep reinforcement learning methods and studying important components to achieve efficient distributed learning,covering single player single agent distributed deep reinforcement learning to the most complex multiple players multiple agents distributed deep reinforcement learning.Furthermore,we review recently released toolboxes that help to realize distributed deep reinforcement learning without many modifications of their non-distributed versions.By analysing their strengths and weaknesses,a multi-player multi-agent distributed deep reinforcement learning toolbox is developed and released,which is further validated on Wargame,a complex environment,showing the usability of the proposed toolbox for multiple players and multiple agents distributed deep reinforcement learning under complex games.Finally,we try to point out challenges and future trends,hoping that this brief review can provide a guide or a spark for researchers who are interested in distributed deep reinforcement learning.

Overhead-free Noise-tolerant Federated Learning: A New Baseline

Federated learning (FL) is a promising decentralized machine learning approach that enables multiple distributed clients to train a model jointly while keeping their data private. However, in real-world scenarios, the supervised training data stored in local clients inevitably suffer from imperfect annotations, resulting in subjective, inconsistent and biased labels. These noisy labels can harm the collaborative aggregation process of FL by inducing inconsistent decision boundaries. Unfortunately, few attempts have been made towards noise-tolerant federated learning, with most of them relying on the strategy of transmitting overhead messages to assist noisy labels detection and correction, which increases the communication burden as well as privacy risks. In this paper, we propose a simple yet effective method for noise-tolerant FL based on the well-established co-training framework. Our method leverages the inherent discrepancy in the learning ability of the local and global models in FL, which can be regarded as two complementary views. By iteratively exchanging samples with their high confident predictions, the two models “teach each other” to suppress the influence of noisy labels. The proposed scheme enjoys the benefit of overhead cost-free and can serve as a robust and efficient baseline for noise-tolerant federated learning. Experimental results demonstrate that our method outperforms existing approaches, highlighting the superiority of our method.

Elastic Optimization for Stragglers in Edge Federated Learning

To fully exploit enormous data generated by intelligent devices in edge computing,edge federated learning(EFL)is envisioned as a promising solution.The distributed collaborative training in EFL deals with delay and privacy issues compared to traditional centralized model training.However,the existence of straggling devices,responding slow to servers,degrades model performance.We consider data heterogeneity from two aspects:high dimensional data generated at edge devices where the number of features is greater than that of observations and the heterogeneity caused by partial device participation.With large number of features,computation overhead on the devices increases,causing edge devices to become stragglers.And incorporation of partial training results causes gradients to be diverged which further exaggerates when more training is performed to reach local optima.In this paper,we introduce elastic optimization methods for stragglers due to data heterogeneity in edge federated learning.Specifically,we define the problem of stragglers in EFL.Then,we formulate an optimization problem to be solved at edge devices.We customize a benchmark algorithm,FedAvg,to obtain a new elastic optimization algorithm(FedEN)which is applied in local training of edge devices.FedEN mitigates stragglers by having a balance between lasso and ridge penalization thereby generating sparse model updates and enforcing parameters as close as to local optima.We have evaluated the proposed model on MNIST and CIFAR-10 datasets.Simulated experiments demonstrate that our approach improves run time training performance by achieving average accuracy with less communication rounds.The results confirm the improved performance of our approach over benchmark algorithms.

MobSafe:Cloud Computing Based Forensic Analysis for Massive Mobile Applications Using Data Mining

With the explosive increase in mobile apps, more and more threats migrate from traditional PC client to mobile device. Compared with traditional Win＋Intel alliance in PC, Android＋ARM alliance dominates in Mobile Internet, the apps replace the PC client software as the major target of malicious usage. In this paper, to improve the security status of current mobile apps, we propose a methodology to evaluate mobile apps based on cloud computing platform and data mining. We also present a prototype system named MobSafe to identify the mobile app＇s virulence or benignancy. Compared with traditional method, such as permission pattern based method, MobSafe combines the dynamic and static analysis methods to comprehensively evaluate an Android app. In the implementation, we adopt Android Security Evaluation Framework （ASEF） and Static Android Analysis Framework （SAAF）, the two representative dynamic and static analysis methods, to evaluate the Android apps and estimate the total time needed to evaluate all the apps stored in one mobile app market. Based on the real trace from a commercial mobile app market called AppChina, we can collect the statistics of the number of active Android apps, the average number apps installed in one Android device, and the expanding ratio of mobile apps. As mobile app market serves as the main line of defence against mobile malwares, our evaluation results show that it is practical to use cloud computing platform and data mining to verify all stored apps routinely to filter out malware apps from mobile app markets. As the future work, MobSafe can extensively use machine learning to conduct automotive forensic analysis of mobile apps based on the generated multifaceted data in this stage.