Text-to-image (T2I) diffusion models have achieved remarkable progress in generating realistic images from textual descriptions. However, ensuring consistent high-quality image generation with complete backgrounds, object appearance, and optimal texture rendering remains challenging. This paper presents a novel fine-grained pixel-level image editing method based on pre-trained diffusion models. The proposed dual-branch architecture, consisting of Guidance and Generation branches, employs U-Net Denoisers and Self-Attention mechanisms. An improved DDIM-like inversion method obtains the latent representation, followed by multiple denoising steps. Cross-branch interactions, such as KV Replacement, Classifier Guidance, and Feature Correspondence, enable precise control while preserving image fidelity. The iterative refinement and reconstruction process facilitates finegrained editing control, supporting attribute modification, image outpainting, style transfer, and face synthesis with Clickand-Drag style editing using masks. Experimental results demonstrate the effectiveness of the proposed approach in enhancing the quality and controllability of T2I-generated images, surpassing existing methods while maintaining attractive computational complexity for practical real-world applications.

Automatic generation of ophthalmic reports holds significant potential to lessen clinicians’ workload, enhance work efficiency, and alleviate the imbalance between clinicians and patients. Recent advancements in multimodal large language models, represented by GPT-4, have demonstrated remarkable performance in the general domain. However, training such models necessitates a substantial amount of paired image-text data, yet paired ophthalmic data is limited, and ophthalmic reports are laden with specialized terminologies, making it challenging to transfer the training paradigm to the ophthalmic domain. In this paper, we propose EyeGraphGPT, a knowledge graph enhanced multimodal large language model for ophthalmic report generation. Specifically, we construct a knowledge graph by leveraging the knowledge from a medical database and expertise from ophthalmic experts to model relationships among ophthalmic diseases, enhancing the model’s focus on key disease information. We then perform relation-aware modal alignment to incorporate knowledge graph features into visual features, and further enhance modality collaboration through visual instruction fine-tuning to adapt the model to the ophthalmic domain. Our experiments on a real-world dataset demonstrates that EyeGraphGPT outperforms previous state-of-the-art models, highlighting its superiority in scenarios with limited medical data and extensive specialized terminologies.

Federated learning (FL) enables collaborative learning across multiple biomedical data silos with multimodal foundation models while preserving privacy. Due to the heterogeneity in data processing and collection methodologies across diverse medical institutions and the varying medical inspections patients undergo, modal heterogeneity exists in practical scenarios, where severe modal heterogeneity may even prevent model training. With privacy considerations, data transfer cannot be permitted, restricting knowledge exchange among different clients. To trickle these issues, we propose a cross-modal prototype imputation method for visual-language understanding (Buffalo) with only a slight increase in communication cost, which can improve the performance of fine-tuning general foundation models for downstream biomedical tasks. We conducted extensive experiments on medical report generation and biomedical visual question-answering tasks. The results demonstrate that Buffalo can fully utilize data from all clients to improve model generalization compared to other modal imputation methods in three modal heterogeneity scenarios, approaching or even surpassing the performance in the ideal scenario without missing modality.

In this paper, the problem of secondary voltage restoration in an islanded microgrid (MG) is considered, in which the actuator of the distributed generators (DGs) may coexist with partial loss of effectiveness (PLOE) fault and bias fault. For each DG, an adaptive event-triggered fault-tolerant (ETFT) control protocol is designed to compensate for the effects of actuator faults in the DGs, thereby restoring the voltage to the reference value. The dependent event triggering mechanism saves the processor’s computational resources. A desirable feature of the protocol proposed in this paper is that the control protocol relies only on relative information between neighboring DG’s, independent of global information about the network graph, fault boundaries, and network scale. It means that the protocol is implemented in a fully distributed framework. The protocol fully applies to the common ETFT consensus control problem of linear multiagent systems (MASs) with actuator faults. Furthermore, comprehensive theoretical arguments for consensus stability and analysis of Zeno behavior ensure the approach’s feasibility. The simulation results verify the effectiveness of the algorithm.Note to Practitioners—This paper aims to propose a fully ETFT control protocol for secondary voltage restoration of an islanded MG. The control protocol consists of a linear term and a nonlinear time that compensates for the multiplicative and additive fault of the actuator. Moreover, DGs’ communication structure and global fault boundaries may be unknown in practice. Hence, adaptive coupling gains that depend only on the sampled relative information of DGs are introduced to estimate the controller gain, thus avoiding global information. A feasible strategy is provided for industrial applications.

This article proposes an observer-based event-driven fault-tolerant (OBEDFT) secondary control strategy for AC microgrids (MGs) to achieve load voltage regulation. First, the input-output feedback linearization method transforms the voltage regulation issue into an output feedback tracking problem for linear multiagent systems (MASs) with nonlinear dynamics. This transformation provides the necessary preprocessing for load voltage regulation. Then, an OBEDFT secondary control protocol that considers full-state immeasurability is proposed. The actuators of distributed generators (DGs) may experience partial loss of effectiveness (PLOE) and bias faults, and these fault parameters may be heterogeneous and time-varying. The protocol introduces adaptive techniques to avoid information related to fault parameters while using event-driven mechanisms to achieve discrete measurements of neighboring DG. Additionally, the protocol uses boundary layers to construct smooth controllers that prevent the chattering effect caused by nonsmooth controllers. Finally, simulation results confirm the effectiveness of this load voltage regulation strategy.

Privacy issues and the cyberattacks are two typical threats in network operation, but the problem considering both of them has not been properly addressed. To fill this gap, this article investigates the privacy-preserving consensus strategy against the false data injection attacks (FDIAs) for secondary control of microgrids. An integral sliding mode observer and its supporting controller are developed to against the FDIAs on local units. Since the sliding motion is independent of the controller signal, the proposed strategy has a natural advantage for the FDIAs on the controller command. Furthermore, an observer-based resilient control strategy is proposed to against the FDIAs on the communication lines. It is worth mentioning that the$H_\infty$performance can be obtained by regarding the consensus errors as disturbance. Moreover, the edge-based privacy-preserving algorithm along with the observed-based resilient strategy is proposed. Finally, some simulation examples are used to verify the theoretical results.

Although privacy-preserving is often considered in energy management and energy scheduling of microgrids (MGs), to the best of authors’ knowledge, there are almost no literatures on distributed secondary control of MGs with privacy-preserving. To fill this gap, this article investigates the privacy-preserving problem for voltage restoration of an ac MG. First, the model of the MG is transformed into a linear multiagent system model by feedback linearization. Furthermore, output mask approach, which avoids divulging the initial condition by inserting a dynamic mask on the exchanged state information, is adopted to make the distributed generator agents exactly converge to the reference value rather than the value with a preset certain convergence error by differential private. Meanwhile, the sliding mode control scheme with simple structure is adopted to accelerate the convergence speed of the masked system. However, it is difficult to design the controller directly because of the strong nonlinearity brought by the mask function. To solve this problem, we first implement the controller for the original system, and then extend it to the masked system. After that, we carry out the consensus analysis of the controlled masked system. Finally, the effectiveness of the proposed control scheme is validated in the MATLAB/Simulink environment.

Multi-label text classification is the machine learning task wherein each document is tagged with multiple labels, and this task is uniquely challenging due to high dimensional features and correlated labels. Such text classifiers need to be regularized to prevent severe over-fitting in the high dimensional space, and they also need to take into account label dependencies in order to make accurate predictions under uncertainty. Many classic multi-label learning algorithms focus on incorporating label dependencies in the model training phase and optimize for the strict set-accuracy measure. We propose a new pipeline which takes such algorithms and improves their F1-performance with careful training regularization and a new prediction strategy based on support inference, calibration and GFM, to the point that classic multi-label models are able to outperform recent sophisticated methods (PDsparse, SPEN) and models (LSF, CFT, CLEMS) designed specifically to be multi-label F-optimal. Beyond performance and practical contributions, we further demonstrate that support inference acts as a strong regularizer on the label prediction structure.

Multi-label classification is an important machine learning task wherein one assigns a subset of candidate labels to an object. In this paper, we propose a new multi-label classification method based on Conditional Bernoulli Mixtures. Our proposed method has several attractive properties: it captures label dependencies; it reduces the multi-label problem to several standard binary and multi-class problems; it subsumes the classic independent binary prediction and power-set subset prediction methods as special cases; and it exhibits accuracy and/or computational complexity advantages over existing approaches. We demonstrate two implementations of our method using logistic regressions and gradient boosted trees, together with a simple training procedure based on Expectation Maximization. We further derive an efficient prediction procedure based on dynamic programming, thus avoiding the cost of examining an exponential number of potential label subsets. Experimental results show the effectiveness of the proposed method against competitive alternatives on benchmark datasets.

Ideal point estimation that estimates legislators' ideological positions and understands their voting behavior has attracted studies from political science and computer science. Typically, a legislator is assigned a global ideal point based on her voting or other social behavior. However, it is quite normal that people may have different positions on different policy dimensions. For example, some people may be more liberal on economic issues while more conservative on cultural issues. In this paper, we propose a novel topic-factorized ideal point estimation model for a legislative voting network in a unified framework. First, we model the ideal points of legislators and bills for each topic instead of assigning them to a global one. Second, the generation of topics are guided by the voting matrix in addition to the text information contained in bills. A unified model that combines voting behavior modeling and topic modeling is presented, and an iterative learning algorithm is proposed to learn the topics of bills as well as the topic-factorized ideal points of legislators and bills. By comparing with the state-of-the-art ideal point estimation models, our method has a much better explanation power in terms of held-out log-likelihood and other measures. Besides, case studies show that the topic-factorized ideal points coincide with human intuition. Finally, we illustrate how to use these topic-factorized ideal points to predict voting results for unseen bills.