Semantic image segmentation typically relies on static pixel- wise classifiers that operate over a fixed category space, making them insensitive to the actual semantic composition of a real-world image. In this work, we propose CaSS, a category-aware semantic segmentation framework that introduces image-level semantic priors to dynamically adapt pixel-level classification. Specifically, a pre-trained vision–language model is employed to infer the set of semantic categories present in an image, which is encoded as a structured category prior. This prior is then used to drive a lightweight dynamic parsing network that generates image-conditioned classifier parameters for pixel- wise segmentation. By explicitly constraining the classifier with category-aware priors, CaSS reduces interference from absent classes and enhances both intra-class consistency and inter-class discriminability. The proposed approach follows a large–small model collaboration paradigm, leveraging the strong semantic understanding of vision–language models while preserving efficient pixel-level inference. Extensive experiments on standard semantic segmentation benchmarks demonstrate that CaSS consistently improves segmentation accuracy over state-of-the-art methods with minimal parameter overhead.

As the Internet of Things (IoT) and artificial intelligence (AI) technologies are rapidly evolving, smart healthcare has emerged as a transformative solution to enhance healthcare quality and optimize resource allocation. This study introduces FEI-Hi, a federated edge intelligence paradigm that integrates edge computing with federated learning (FL) to enable secure and efficient medical data processing. FEI-Hi comprises three principal layers: FL layer, which facilitates cross-device collaborative training through encrypted model updates; aggregation layer, which refines the global model by consolidating updates; and edge layer, which performs local data processing and model inference. FEI-Hi leverages distributed intelligent computation, model parameter compression, and efficient node clustering to enhance the accuracy and efficiency of medical data processing significantly. By employing Wasserstein distance for clustering and parameter selection, FEI-Hi ensures model convergence and stability. Experimental results on multiple medical datasets demonstrate a 30% improvement in the model training speed and an F1-score exceeding 90%, surpassing the state-of-the-art (SOTA) benchmarks in model parameter transfer efficiency, training speed, and accuracy.

The contactless interaction approach enables robots to identify object attributes via sensors and advanced perceptual technologies, eliminating the need for physical contact. This approach enhances interaction flexibility and safety while offering a richer, more intuitive human-robot experience. Current research focuses on cross-modal retrieval through non-contact feature extraction, enabling material attribute estimation and advancing object perception. However, existing cross-modal material retrieval methods overlook learning common attributes shared within the same material type. To address this, we propose Materials Perception with Self-paced Prototype contrastive Learning (MPSPL). First, attribute prototype contrastive learning extracts shared characteristics within the same material type, aggregates similar samples, and enhances the perception of deep model material attributes. Second, a self-paced learning strategy dynamically adjusts the contrastive loss temperature coefficient, guiding the model to discern discriminative features across materials and progressively recognize inherent attributes. Finally, extensive experiments validate our method under both known and unknown material category conditions. Experiments demonstrate the superiority of our proposed method over state-of-the-art approaches.

The multicloud environments (MCE) represent a novel paradigm encompassing multiple infrastructure as a service (IaaS) providers, enabling users to tailor and optimize cloud services according to their specific requirements. This approach effectively addresses the limitations of a single cloud environment (SCE) regarding technical constraints, geographical coverage deficiencies, and cost-effectiveness concerns while catering to the increasingly diverse and expanding user demands. In MCE, users must employ appropriate strategies to efficiently allocate diverse tasks across multiple cloud service providers (CSPs) by leveraging the best available resources. Traditional scheduling algorithms are inadequate for addressing the complexities of such MCE. This study introduces a framework for the task scheduling procedure in MCE, treating independent task scheduling as a Markov decision process (MDP). We propose a novel agent environment framework that is designed based on the distinctive characteristics of MCE and enables independent task scheduling. Furthermore, we propose a task scheduling algorithm for MCE based on deep reinforcement learning (DRL) to optimize cost and makespan according to diverse user requirements. The simulation experiments are conducted using both simulated datasets and real-world datasets, demonstrating that our proposed algorithm surpasses the other five algorithms in terms of cost minimization and makespan optimization.

The encoder-decoder architecture is a prevailing paradigm for semantic segmentation. It has been discovered that aggregation of multi-stage encoder features plays a significant role in capturing discriminative pixel representation. In this work, we rethink feature reconstruction for scale alignment of multi-stage pyramidal features and treat it as a Query Update (Q-UP) task. Pixel-wise affinity scores are calculated between the high-resolution query map and low-resolution feature map to dynamically broadcast low-resolution pixel features to match a higher resolution. Unlike prior works (e.g. bilinear interpolation) that only exploit sub-pixel neighborhoods, Q-UP samples contextual information within a global receptive field via a data-dependent manner. To alleviate intra-category feature variance, we substitute source pixel features for feature reconstruction with their corresponding category prototype that is assessed by averaging all pixel features belonging to that category. Besides, a memory module is proposed to explore the capacity of category prototypes at the dataset level. We refer to the method as Category Prototype Transformer (CPT). We conduct extensive experiments on popular benchmarks. Integrating CPT into a feature pyramid structure exhibits superior performance for semantic segmentation even with low-resolution feature maps, e.g. 1/32 of the input size, significantly reducing computational complexity. Specifically, the proposed method obtains a compelling 55.5% mIoU with greatly reduced model parameters and computations on the challenging ADE20K dataset.

Data stream classification is widely used in Internet of Things (IoT) scenarios such as health monitoring, anomaly detection and online diagnosis. Due to the continuous data stream changing dynamically over time, it is impossible to classify all the data simultaneously. Moreover, labeling each sample in practical data stream applications is time-and resource-consuming. The realistic situation is that only a few instances in a data stream are labeled. Therefore, classifying data streams with limited labels has become challenging in IoT scenarios. In this paper, we propose an incremental dynamic weighted semi-supervised method for classifying IoT data streams. Considering the dynamics and continuity in data streams, we use a chunk-based approach to learn the features in the data stream and assign weights to the classifier dynamically. Moreover, we deploy incremental learning methods to continuously learn from the sampled labeled data stream to update the classifier model, which can take advantage of newly incoming labeled data to improve learning performance. Experimental evaluations on seven IoT datasets show that the proposed method outperforms semi-supervised methods in accuracy, precision, and geometric mean (Gmean) by 10% and 5% over supervised methods, respectively.

Anomaly detection has become a key focus in maintaining the stability and reliability of the cloud environment. Although with excellent feature extraction ability, deep learning-based anomaly detection methods entail a time-consuming training process. Broad learning system (BLS) provides an alternative supervised way for efficient training. However, due to the imbalance of the collected cloud computing data in which anomaly accounts for a low proportion, sufficient feature extraction from anomaly behaviors with BLS becomes a challenge. Moreover, the input generation of BLS only considers the independence of data, and the generalization of BLS in the correlation modeling of cloud computing data is limited. To tackle the above issues, we introduce an effective anomaly detector, CatBRS, an improved BLS with rebalance operations. Initially, we employ a hybrid resampling method of SMOTE-Tomek to mitigate data imbalance, retain non-synthetic samples for training, and involve synthetic samples in the input generation later. Subsequently, we extend BLS by refining the process of input generation. This enhanced system employs a simple recurrent architecture to model temporal dynamics. Additionally, it integrates an autoencoder-based model with metric learning to obtain category-constrained discriminant features. The improvement in BLS facilitates more comprehensive feature extraction. Finally, extensive experiments are conducted to evaluate the performance of CatBRS on four benchmark datasets. CatBRS shows improvements of up to 3.81% in AUC and 6.09% in F1 compared to suboptimal baseline methods with a low training cost.

Recently, the multi-cloud environment (MCE) has increasingly become the preferred choice of users. As with the cloud environment, efficient workflow scheduling in a MCE remains crucial for identifying the cost efficiency and overall performance of the MCE. In MCE, the resources exhibit heterogeneity, complexity, and dynamism. Simultaneously, the intricate inter-task dependencies among workflow tasks, diverse Quality of Service (QoS) metrics for users, and multiple cloud service providers’ (CSPs) billing mechanisms significantly amplify the workflow scheduling challenge. Motivated by the application of reinforcement learning (RL) in workflow scheduling in a cloud environment, this paper proposes a scheduling algorithm that takes advantage of the asynchronous advantage actor–critic algorithm (A3C) to balance cost, makespan and resource utilization in workflow scheduling in a MCE. By analyzing the elements in the MCE, we design and define multiple agents in the MCE, and each cloud service provider will have an agent to record the state and update the local parameters. For the workflow task submitted by the user, the action is selected according to the initialization policy and submitted to the scheduling action to allocate the task to a designated virtual machine in the MCE so that each agent can more clearly perceive the environment change and adapt to the MCE. In contrast to the traditional A3C algorithm, we design a new critic network according to the data characteristics of real-world scientific workflows so that each agent is more suitable for real-world scientific workflow data. Through multiple sets of simulation experiments, the workflow scheduling algorithm based on the A3C algorithm in the MCE (MCWS-A3C) was compared with three benchmark methods. The experimental results show that the proposed method has better advantages than other methods in terms of cost, makespan, and resource utilization . Specifically, on the Montage_100 dataset, the average cost was reduced by 55.12% compared to other methods. The pioneering introduction of the A3C algorithm that adapts to the dynamic environment into the MCE brings more possibilities to address the issue of workflow scheduling in the MCE.

The anomaly detection techniques for cloud service focus on alerting the operation engineers about the anomalous running state. However, their shortcoming of anomaly interpretability is an obstacle to understanding and further removing the anomalies. To overcome the abovementioned challenge, we propose a tree-like attention-based detection framework CauseFormer that provides both the metric and sample interpretations. Firstly, we develop stepwise attention based on the multi-head attention mechanism, which imitates the rule-based tree formation process. This network block extracts the higher-order features and generates the metric contribution that can be regarded as metric interpretation. Meanwhile, we design the hyper-circle loss function rather than cross-entropy-based approaches to optimize the representation. Then we introduce the majority voting rule into the classifier. This neighbor classification criterion raises the alarms of anomalies and achieves the sample interpretation. Finally, we conduct extensive experiments in four datasets collected from cloud application cases. The experimental results reveal the superiority of CauseFormer in improving detection accuracy and embodying practical interpretability.

Anomaly detection is indispensable for achieving higher availability and reliability in the cloud computing. The traditional autoencoder-based method only models the historical normal samples and then identifies the current online anomaly samples by the fixed threshold of anomaly score. Although more advances have been made in recent years, two main challenges remain: (i) ignoring the historical anomaly samples, (ii) poor self-adaptive ability for online detection. To address the above challenges, we propose a unified detector, namely BroadCAE, which integrates autoencoder with contrastive learning and broad network. Specifically, the reconstruction loss is first replaced by contrastive loss, which equally formulates both normal and anomaly samples. These samples belonging to the same class become closer in a lower-dimensional space. Conversely, different classes of samples are far away from each other. Next, we apply the anomaly-score-based pseudo thresholds to train the dynamic threshold selection, which generates the threshold according to the coming sample. The broad network in dynamic threshold selection takes the place of the deep network, which overcomes catastrophic forgetting and adapts to new online samples. Finally, validation experiments are conducted on four benchmark datasets. Our BroadCAE outperforms the comparative baseline methods by averaging over 4% of the f1-score.

Current scene segmentation methods suffer from cumbersome model structures and high computational complexity, impeding their applications to real-world scenarios that require real-time processing. This paper proposes a novel Efficient Pyramid Representation Network (EPRNet), which strikes an innovative record on segmentation accuracy, model lightness and inference efficiency. Unlike existing methods delivering transfer learning based on pixel features of limited receptive fields encoded by shallow image classification backbones, EPRNet distributes multi-scale representations throughout the feature encoding flow to quickly enlarge and enrich receptive fields. Specifically, we introduce an extremely lightweight and efficient Multi-scale Processing Unit (MPU) that encodes multi-scale features through parallel convolutions of different kernels. By combining MPU and residual learning, we propose a core Pyramid Representation Module (PRM) to correctly acquire and aggregate region-based contexts in both shallow and deep layers. In this way, EPRNet can encode discriminative and comprehensive representations of multi-scale objects with a compact structure. We conduct extensive experiments on Cityscapes and CamVid datasets, demonstrating the superiority. Without any extra and coarse labeled data, EPRNet obtains mIoU 73.9% on the Cityscapes test set with only 0.9 million parameters at a speed of 42 FPS.

Semantic segmentation plays a critical role in scene understanding for self-driving vehicles. A line of efforts has proven that global context matters in urban scene segmentation due to massive scale changes. However, we find that existing methods suffer from local ambiguities when dissipating continuous local context, i.e. scrambling to a huge receptive field of global cues by coarse pooling. To this end, this paper proposes a new Context Aggregation Module (CAM) that consists of two primary components: context encoding using no coarse pooling but encoder-decoders with appropriate sampling scales and gated fusion that extends gate attention mechanism to balance different-scale context during feature fusion. Weeding out coarse pooling and applying the encoder-decoder inherits the merits of exploring global context while avoiding the drawback of losing local contextual continuity. We then construct a Context Aggregation Network (CANet) and conduct extensive evaluations on challenging autonomous driving benchmarks of Cityscapes, CamVid and BDD100K. Consistently improved results evidence the effectiveness. Notably, we attain competitive mIoU 82.7% on Cityscapes and optimal mIoU 80.5% on CamVid.