MOTIVATION: Accurate prediction of adverse drug reactions (ADRs) is essential for drug safety surveillance, and recent advances in machine learning with heterogeneous biomedical information have improved predictive performance. However, two challenges remain: current methods often learn inadequate ADR representations that fail to capture dependencies among ADRs, and generalize poorly to novel drugs. RESULTS: To obtain informative ADR embeddings, we construct a multi-source, multi-relational ADR graph that integrates hierarchical structure and empirical ADR co-occurrence, and apply a relational graph convolutional network (R-GCN) to learn relation-aware ADR representations. To enhance generalization to novel drugs, we exploit the hierarchical structure of the Anatomical Therapeutic Chemical (ATC) classification to link drugs via shared higher-level categories for effective knowledge transfer and model these relations with an R-GCN. We further introduce a Conditional Domain Adversarial Network (CDAN) to reduce distribution shifts between known and novel drugs by aligning features conditioned on predicted ADR labels, learning domain-invariant yet task-relevant representations. Additionally, to exploit similar ADR patterns among related drugs, we introduce a dual-branch mixture-of-experts (Dual-MoE) module where each expert captures ADR commonalities within a drug category in one branch, while a separate branch models global patterns. Extensive experiments show that our method consistently outperforms seven baselines, achieving F1 improvements of 4.3% and 4.7% over the best baseline on two datasets, respectively, with more balanced precision-recall trade-offs. It also improves AUC on uncommon ADRs by 7% more than on common ADRs, and remains more robust under data sparsity, with more gradual performance degradation as training data decreases. AVAILABILITY AND IMPLEMENTATION: The code of our model is available at https://github.com/fzsdb/Knowledge-guided-ADR-prediction.git.

Although information interaction is essential for achieving group goals of distributed systems, the information interaction consumes lots of communication resources, and also leads to the privacy disclosure of individual information. Therefore, how to design an efficient and secure distributed information interaction algorithm naturally becomes a crucial issue. In this article, we propose a differential privacy consensus algorithm based on an event-triggered communication mechanism for current sharing and voltage regulation in DC Microgrids. The main idea is to, at each event-triggered instant, first release some sporadic current data, and then mask the data with additive Laplace noises with an invariant variance. By virtue of a stochastic approximation technique, time-varying control gains are designed to compensate for the adverse effect of the stationary noises. As a result, not only the execution efficiency of the network is improved, but the privacy of initial currents can be well protected. The advantage of the proposed algorithm lies in its ability to prevent the gradual leakage of privacy caused by noise attenuation. That is, it provides stronger privacy protection than the commonly adopted approach using exponentially decaying noise. Furthermore, we carry out rigorous convergence analysis for current sharing and average bus voltage regulation, and also evaluate the level of differential privacy achieved. Finally, the effectiveness of the proposed algorithms is validated by simulation results from a detailed switch-level microgrid model.

In this article, we consider the problem of privacy preservation in consensus control for second-order integrator multiagent systems (MASs). Specifically, we consider the setting where the initial position and velocity of each legitimate agent are both private, an internal or external adversary wants to identify them based on the information it obtains. To deal with this scenario, we propose a privacy preservation algorithm based on a position and velocity simultaneous perturbation technique. To be specific, our algorithm consists of a collaborative scrambling phase and a convergence phase. In the scrambling phase, each agent is required to produce two sets of edge-based perturbation signals that are, respectively, imposed on the local position and velocity signals before transmission, with the purpose of preserving privacy; in the convergence phase, each agent updates its state per a normal rule, aiming to achieving accurate consensus. Also, we establish a system-theoretic framework to analyze privacy performance by examining the indistinguishability of private values' arbitrary variations to adversaries, and further show that, an internal adversary cannot infer the privacy of a legitimate agent provided it has at least one legitimate in-neighbor or out-neighbor, and the privacy is leaked out once that agent exclusively connects to the internal adversary in bidirectional way. As for external eavesdroppers, they can never infer any agent's privacy if the gain parameters in the scrambling phase are not accessible to them. Finally, two simulation examples illustrate the validity of the proposed approach.

The increasing number of devices and frequent interactions of agents from networked multiagent systems (MASs) exacerbate the risks of potential cyber attacks, especially the different point attacks and multiple pattern attacks. This article considers the output formation-containment problem for MASs under multipoint multipattern false data injection (FDI) attacks. The multipoint describes the attacks simultaneously occurring on the sensors, actuators, and communication channels; the multipattern captures that sensor and actuator attack signals are both continuous deterministic variables, and the communication channel attack signals are intermittent random variables, obeying the Bernoulli distribution. For such compromised MASs, a novel hybrid protocol is proposed, which integrates a state observer, an attack estimator, an impulsive interactor and a compensation controller. Thereinto, the state observer and the attack estimator are constructed to recover the unmeasured system states and the unknown FDI attack signals, respectively; the impulsive interactor is designed to guarantee that the neighbor's signals are transmitted only at impulsive instants, and meanwhile the channel attacks are randomly launched; using the recovered signals, the compensation controller is devised to alleviate the effect of attacks. A sufficient condition is identified, under which the output formation containment is achieved with cooperative uniform ultimate boundedness (UUB). Finally, simulation results are carried out to validate the effectiveness and advantages of the proposed approach.

The relative state between neighbors represents the difference of two connected agents' states, and it possesses specific physical meanings in practice. Under this background, the saturation constraints in the relative state inevitably occur. This article studies the consensus problems under the relative state saturation constraints. Novel adaptive proportional-integral (PI) protocols are designed to solve the constrained consensus problem. Specifically, the adaptive coupling weights and the saturation functions are embedded into the proposed protocols, and the former can render the protocols independent of any global topology graph information, while the latter can confine the relative state to stay in its constrained set. Sufficient conditions are identified under which the constrained consensus can be achieved. Considering that the solution matrix is required to be diagonally dominant, an iterative learning-based heuristic algorithm is proposed to seek the diagonally dominant positive-definite solution matrix. For the special case that the input matrix is row full rank, more stringent saturation functions are constructed, and it not only achieves the constrained consensus but also realizes the nonovershoot and shorter settling time associated with edge states. Besides, this result can be applied to preserve connectivity of the communication network. The theoretical analyses are validated by a simulation example.

Relative states between neighbors are ubiquitous in large-scale systems, and the relative state saturations need to be considered when designing controllers based on this relative state information. This article investigates the consensus control of the delayed multiagent systems under the relative state saturations. By means of an incidence matrix, the consensus under the relative state saturations is transformed into the stability of edge dynamics operating on constrained sets. To enable the edge states to stay within the constrained sets, a novel nonlinear protocol is designed by embedding an elaborate saturation function. Sufficient conditions are identified, under which not only consensus is achieved but also relative state saturations do not occur. Moreover, this analytical result can address the consensus problem while preserving connectivity, within the limited communication range framework. Simulations illustrate the theoretical results.

This article investigates the consensus tracking of exponentially unstable multiagent systems with time-varying input delay. The truncated predictor feedback approach is utilized for designing delay-dependent state and output feedback protocols. And the explicit conditions that can realize consensus tracking are established in terms of parametric Lyapunov equations and scalar inequalities. Besides, the protocol design algorithms under the maximum allowable delay and the maximum convergence rate are, respectively, provided. Compared with the existing results, the salient characteristic of the current results is to reveal the quantitative relationship among the time delay, unstable plant, network topologies, and the convergence rate. Specifically, for achieving the consensus tracking, the synchronization force from network connectivity needs to dominate the anti-synchronization force from unstable open-loop poles and input delays; the maximum allowable input delay is inversely proportional to the sum of the unstable open-loop poles; the convergence rate becomes smaller as the input delay bounds or/and the sum of the unstable poles in plant increase. Numerical simulations confirm the effectiveness of the proposed theoretical design.

Within the multiagent systems framework, the relative states between neighbors can be acquired by some on-board sensors, and then the relative state saturations inevitably occur due to the limited sensing capabilities. This paper investigates the consensus problem of nonlinear multiagent systems subject to the relative state saturations. Utilizing the incidence matrix and the edge Laplacian, the consensus problem of nonlinear multiagent systems with the relative state saturations can be cast into the stabilization problem of edge dynamics operating on the constrained set. Three types of protocols, namely continuous, intermittent, and adaptive state feedback protocols are, respectively, proposed for achieving the constrained consensus, and meanwhile yielding consensus values. A consensus analysis is provided by virtue of state saturation theory, switched system theory, adaptive theory, and Lyapunov stability theory. Output feedback protocol is also designed. Finally, the obtained results are applied to connectivity preservation for first-order nonlinear multiagent systems, despite the presence of limited communication range and input constraints. The theoretical results are validated by two simulation examples.

Within the framework of multiagent systems, relative information can be directly acquired by vehicle-mounted sensors and the relative information constraints inevitably occur due to limited sensing capabilities. This paper investigates observer-based consensus of nonlinear multiagent systems subject to relative state estimate constraints. Each agent's state is constructed via a state observer, and the relative state estimate is assumed to be confined into a hypercube. In virtue of the edge Laplacian, the consensus problem of nonlinear multiagent systems under this constraint is converted into the stabilization problem of edge dynamics operating on the constrained set. Observer-based intermittent protocol and adaptive protocol are, respectively, designed for achieving consensus. A convergence analysis is provided with the help of state saturation theory, switched system theory and adaptive theory. Finally, the results on consensus with relate state estimate constraints are applied into the consensus problem while preserving the connectedness, and are validated by a simulation example.

This paper investigates leaderless consensus and leader-following consensus of multiagent systems with Lipschitz nonlinearity and input delay. For such systems, the observer-based truncated prediction feedback protocols are designed via dropping the distributed term and remaining the exponential term of the solution of the system equation over the delay period. Using Lyapunov-Krasovskii functional approach, two sufficient criteria are, respectively, established for achieving leaderless and leader-following consensus. The observer and controller gains are then obtained by means of an iterative linear matrix inequality procedure. A numerical simulation verifies the effectiveness of the proposed control design approach.

For leader-following multiagent systems with input saturation, the existing protocols use a low gain feedback approach to achieve semi-global consensus. The main drawback of this approach is the ineffective utilization of the actuator potential, resulting in bad performance. To improve the transient performance of the consensus tracking, this paper proposes a gain scheduled approach for multiagent systems subject to the saturator saturations. A novel kind of scheduler-based protocols are proposed, which consists of state feedback controllers with time-varying gain and parameter schedulers. The role of the controllers is to achieve the consensus tracking, while the schedulers can accelerate this consensus progress by enlarging the gain parameter. To remove the dependence of the schedulers on global information, a minimum-value-based consensus algorithm is put forward, with idea of driving all values of agents throughout the network to their minimum value. Its implementation is guaranteed by the network-topology connectivity. Finally, our approach is further extended to the case where the leader's control input is nonzero, time-varying, and bounded. The discontinuous protocol and its continuous approximation counterpart are designed, yielding the exactand quasi-consensus tracking, respectively. Simulation results verify the theoretical analysis.

This paper addresses the robust consensus reliable control problem against actuator faults for linear parameter-varying multiagent systems. First, the actuator faults are modeled via a polytopic uncertainty method. Second, a distributed observer is designed for the single agent by sharing the communication network sensors to estimate the state information. Then by these estimated information, a robust consensus reliable control protocol against actuator faults is obtained and desired disturbance rejection performance can be guaranteed by this proposed protocol. Third, the nonconvexity conditions of the consensus control protocol can be translated into an linear matrix inequality optimization problem via simple matrix calculation. Finally, the effectiveness of the proposed reliable controller scheme is illustrated by two examples.