Abstract Personalized federated learning aims to deliver customized models for clients with heterogeneous data distributions while preserving data privacy. In this work, we propose personalized federated learning with feature complementarity and temporal regularization (pFedCT), a novel PFL framework where each client adopts a broad learning system (BLS) as the local model. Indeed, the existing BLS-based FL methods have been successful in enhancing communication efficiency, privacy preservation, and Byzantine robustness, but they often fail to address the critical issue of personalization. To achieve effective personalization, pFedCT introduces two key components. First, we extract feature node matrices from each client’s BLS model, which capture the internal feature representations of local data. By comparing the principal subspaces of these matrices, we assess the complementarity between clients’ learned representations. This enables each client to collaborate with others who provide diverse and informative knowledge. Second, we incorporate a temporal regularization mechanism to stabilize the collaboration structure across communication rounds, mitigating performance degradation caused by fluctuating local updates and data shifts. Extensive experiments on Modified National Institute of Standards and Technology (MNIST) and Fashion-MNIST datasets under both independent and identically distributed (IID) and non-IID settings demonstrate that pFedCT consistently outperforms existing BLS-based baselines and achieves competitive accuracy compared to deep neural network-based methods, while maintaining superior efficiency in communication and computation.

Federated Learning (FL) enables model sharing between clients and a server instead of data transfer, enhancing user privacy. In this study, we employ an asynchronous training approach, allowing the server to proceed without waiting for crashed clients, thus improving round efficiency. Besides, because there is highly heterogeneous data distribution between clients, we propose a dynamic personalized federated learning approach, which helps the global model better fit each client’s unique data distribution. In addition, we also consider that the outdated models can lower the global model’s accuracy because they fail to capture the latest user preferences. we have designed a switchable aggregation algorithm. If there is a large data distribution difference between the latest updated clients and other outdated clients, which means the users’ preferences have changed a lot. To solve this challenge, when aggregating, their model version will be considered. If there is a little difference, which means the users’ preferences haven’t changed a lot, we allow all models in the cache to participate in this round of aggregation regardless of their model version, which greatly improves the efficiency of the aggregation. In this way, the global model will be tilted towards clients’ model with the latest data. Ultimately, PFL-SA demonstrates higher accuracy and reduced communication overhead compared to the other federated learning methods, as evidenced by the experimental results.

At present, the dynamic encoding–decoding scheme is utilized in the impulsive consensus control of multiagent systems (MASs) to solve the limited bandwidth problem. However, the unknown nonlinear dynamics and deception attacks will generate some uncertainties inevitably in the encoding–decoding, which may cause the quantizer saturation and then influence the consensus performance. Therefore, the impulsive consensus control problem of uncertain nonlinear MASs based on encoding–decoding under deception attacks is investigated in this article. To address the system uncertainty, an adaptive algorithm with neural networks is designed. Under the proposed estimator for each follower, the designed adaptive law for every follower only needs the information from its own sensor and estimator instead of the quantized information from its neighbors. Then a more general scenario of uncertain deception attack is considered, in which the uncertain deception attacks can occur in the both parts of hybrid impulsive control protocol. An attack observer is introduced to handle this more complex attack by compensating impacts of deception attacks. Next the sufficient conditions for secure consensus of MASs with limited bandwidth are derived. Moreover, although some uncertainties exist in the encoding–decoding, the quantizer saturation can be eliminated by adjusting the parameters of controller. Finally, the validity of given theorems is demonstrated by the simulation experiments.

In response to the growing demand for clients' privacy protection, federated learning framework often needs appropriate privacy protection to protect clients' privacy better, such as the client uploading a blinded local model instead of the original real local model to the server, which makes the real value of the local model unobservable to the server. Although the privacy is protected, it will to be a huge challenge to the server to distinguish clients which are Byzantine clients. We propose a federated learning framework based on broad learning that can simultaneously achieve protection of clients' privacy and robustness against Byzantine attacks, i.e. FBL-BP. We apply differential privacy techniques to perturb the local models of the clients, which can protect clients' privacy. The server receives the perturbed local models and then guarantees the Byzantine-resilience of global model through an outlier removal mechanism based on cosine similarity. Finally, experimental results show that FBL-BP has significant Byzantine robustness and satisfactory accuracy, and possesses less time consumption than traditional methods.

In this article, the event-triggered bipartite consensus problem for stochastic nonlinear multiagent systems (MASs) with unknown dead-zone input under the prescribed performance is studied. To surmount the influence of the dead-zone input, the dead-zone model is transformed into a linear term and a disturbance term. Meanwhile, the prescribed tracking performance is realized by developing a speed function, which means that all tracking errors of MASs can converge to a predefined set in a given finite time. Moreover, the unknown nonlinear dynamics are approximated by fuzzy-logic systems. By combining the dynamic surface approach and the Lyapunov stability theory, we design an adaptive event-triggered control algorithm, such that the bipartite consensus problem of stochastic nonlinear MASs can be achieved, and all signals are semiglobally uniformly ultimately bounded in probability of the closed-loop systems. Finally, simulation examples are proposed to verify the feasibility of the algorithm.

This paper solves the adaptive event-triggered consensus problem for nonlinear multi-agent systems with state time delay and unknown external disturbance. First, the disturbance is estimated by designing a novel adaptive disturbance observer. Next, Lyapunov-Krasovskii functional is utilized to deal with the state delay. Finally, in accordance with the designed event-triggered condition, an adaptive controller is proposed, in which neural networks are applied to approximate the unknown nonlinearity. According to Lyapunov stability theory, it can be proved that nonlinear multi-agent systems can reach consensus under the proposed controller. The simulation validates the theoretical results.

Broad Learning System (BLS) is effective and efficient in dealing with various machine learning problems. When constructing a BLS based classifier with domain adaption capability, the mapped features and enhancement nodes are determined by random parameters. Thus, there will be discrepancy between the hidden layer features of the source domain and target domain. Therefore, we propose a new semi-supervised classifier with domain adaption capability based on BLS. Firstly, in order to reduce the discrepancy between the hidden layer features of domains, our method aligns the second-order statistics of mapped features due to the fact that enhancement nodes are generated by mapped features. Further, when learning the classifier through the aligned features, we embed balanced distribution adaptation to improve domain adaption capability of the classifier. Experiments on benchmark datasets demonstrate our method has better classification accuracy than some existences.

In this article, we propose the swarm control for a self-organized system with fixed and switching topology, which can realize aggregation, dispersion, or switching formation when swarm moves. The self-organized system can automatically construct the communication topology for intelligent units in swarm. Swarm control can realize aggregation and dispersion of intelligent units based on its communication topology when swarm moves. The proposed swarm control, in which distances between the related intelligent units are time varying, is different from traditional swarm consensus or swarm formation maintenance. To design swarm control, we define the normalization adjacency matrix and normalization degree matrix based on communication topology. The communication topology is automatically generated based on relation-invariable persistent formation. Depending on whether the communication topology changes or not, the swarm control can be classified as fixed topology and switching topology. Then, the swarm control with fixed and switching topology is designed and analyzed, respectively. The swarm control can realize stability asymptotically when topology is fixed and realize stability in finite time when topology is switched. The simulation results show that the proposed approaches are effective.

This paper focuses on the leader-following consensus control problem of multiagent systems in random vibration environment. The Itô stochastic systems with heterogeneous unknown dynamics and external disturbances are established to describe the agents in random vibration environment. The fuzzy logic systems are applied to approximate the unknown nonlinear dynamics, and one adaptive parameter is designed to decay the effect of external disturbances. We present a new distributed consensus controller for each follower agent only based on local information that is measured or received from its neighbors and itself. Under the consensus controller, we prove that all the follower agents can keep consensus with the leader, even though only a very small part of follower agents can measure or receive the state information of the leader. Furthermore, the states of all the follower agents are bounded in probability. Finally, the simulation results are provided to illustrate the effectiveness of the designed algorithm.

In this paper, the consensus tracking control problem of second-order multiagent systems with unknown nonlinear dynamics, immeasurable states, and disturbances is investigated. The nonlinear dynamics in multiagent systems do not satisfy the matched condition. In this paper, fuzzy logic system is introduced to approximate the unknown nonlinear dynamics, and adaptive high-gain observer is designed to estimate the unmeasured states. Based on backstepping approach and Lyapunov theory, a new adaptive fuzzy distributed controller is proposed for each agent only using the information of itself and its neighbors. Then the consensus tracking is achieved under the designed distributed controller. Moreover, it is proved that all the signals in the multiagent systems are semiglobally uniformly ultimately bounded, and the consensus tracking error converges to a small neighborhood of the origin that can be designed as small as possible. Finally, the simulation result illustrates the effectiveness of the designed controller.

A novel decentralized control approach for second-order uncertain nonlinear multi-agent systems is presented. The communication topology of the multi-agent system that describes completely unknown nonlinear dynamics and unmeasured states is described by a directed graph. The proposed decentralized control algorithm is developed based on high-gain observer theory and fuzzy adaptive control algorithm. The high-gain observer is introduced to estimate the agents' unmeasured states. The fuzzy logic systems are used as the approximator to deal with the nonlinear unknown dynamics. By the Lyapunov theory and consensus analysis, we can prove the consensus errors and the observer errors can be reduced as small as desired by choosing the appropriate design parameters. Finally, the effectiveness of the proposed approach is illustrated by the simulation example.