Search

Data-driven machine learning techniques have been advocated for signal detection in complex wireless environments. However, when applied to wideband spectrum sensing scenarios, they face practical challenges including very large data dimensionality, insufficient training data, and implicit inter-band dependencies. Current literature focuses on deep convolutional models, whose inherent model structure is not well suited for representing the diverse spectrum occupancy patterns of practical wideband networks, causing inefficient performance-complexity tradeoff and excessive sensing time. To address these issues, this paper develops a novel Spectrum Transformer with multi-task learning for wideband spectrum sensing at high sample efficiency. Empowered by the multi-head self-attention mechanism, the transformer architecture is designed to effectively learn both the inner-band spectral features and the inter-band spectrum occupancy correlations in the wideband regime. Simulations show that the proposed Spectrum Transformer outperforms the existing methods based on convolutional neural networks especially in the small-data case, by achieving higher sensing accuracy with an 89% reduction in model complexity.

Non-contact heart rate estimation using remote photoplethysmography (rPPG) has shown great potential in many applications and achieved creditable results in constrained scenarios. However, practical applications require results to be accurate even under complex environment. In this paper, we propose a novel video embedding method that embeds each facial video into a feature map referred to as Multi-scale Adaptive Spatial and Temporal Map with Overlap (MAST_Mop), which contains not only vital information but also surrounding information, which acts as the mirror to figure out the homogeneous perturbations imposed on foreground and background simultaneously, such as illumination instability. Correspondingly, we propose a two-stream Transformer to map the MAST_Mop into heart rate (HR), where one stream follows the pulse signal in the facial area while the other figures out the perturbation signal from the surrounding region to enable feature-level subtraction between the two channels. Due to the context-aware overall feature embedding, the proposed approach outperforms all current state-of-the-art methods on two public datasets MAHNOB-HCI and VIPL-HR.

Millions of patients suffer from rare diseases around the world. However, the samples of rare diseases are much smaller than those of common diseases. Hospitals are usually reluctant to share patient information for data fusion due to the sensitivity of medical data. These challenges make it difficult for traditional AI models to extract rare disease features for disease prediction. In this paper, we propose a Dynamic Federated Meta-Learning (DFML) approach to improve rare disease prediction. We design an Inaccuracy-Focused Meta-Learning (IFML) approach that dynamically adjusts the attention to different tasks according to the accuracy of base learners. Additionally, a dynamic weight-based fusion strategy is proposed to further improve federated learning, which dynamically selects clients based on the accuracy of each local model. Experiments on two public datasets show that our approach outperforms the original federated meta-learning algorithm in accuracy and speed with as few as five shots. The average prediction accuracy of the proposed model is improved by 13.28% compared with each hospital's local model.

This letter is a brief summary of a series of IEEE TIV's decentralized and hybrid workshops (DHWs) on Federated Intelligence for Intelligent Vehicles. The discussed results are: 1) Different scales of large models (LMs) can be federated and deployed on IVs, and three types of federated collaboration between large and small models can be adopted for IVs. 2) Federated fine-tuning of LMs is beneficial for IVs data security. 3) The sustainability of IVs can be improved through optimizing existing models and continuous learning using federated intelligence. 4) LM-enhanced knowledge can make IVs smarter.

Automatic generation of medical reports can provide diagnostic assistance to doctors and reduce their workload. To improve the quality of the generated medical reports, injecting auxiliary information through knowledge graphs or templates into the model is widely adopted in previous methods. However, they suffer from two problems: 1) The injected external information is limited in amount and difficult to adequately meet the information needs of medical report generation in content. 2) The injected external information increases the complexity of model and is hard to be reasonably integrated into the generation process of medical reports. Therefore, we propose an Information Calibrated Transformer (ICT) to address the above issues. First, we design a Precursor-information Enhancement Module (PEM), which can effectively extract numerous inter-intra report features from the datasets as the auxiliary information without external injection. And the auxiliary information can be dynamically updated with the training process. Secondly, a combination mode, which consists of PEM and our proposed Information Calibration Attention Module (ICA), is designed and embedded into ICT. In this method, the auxiliary information extracted from PEM is flexibly injected into ICT and the increment of model parameters is small. The comprehensive evaluations validate that the ICT is not only superior to previous methods in the X-Ray datasets, IU-X-Ray and MIMIC-CXR, but also successfully be extended to a CT COVID-19 dataset COV-CTR.

Face inpainting aims at plausibly predicting missing pixels of face images within a corrupted region. Most existing methods rely on generative models learning a face image distribution from a big dataset, which produces uncontrollable results, especially with large-scale missing regions. To introduce strong control for face inpainting, we propose a novel reference-guided face inpainting method that fills the large-scale missing region with identity and texture control guided by a reference face image. However, generating high-quality results under imposing two control signals is challenging. To tackle such difficulty, we propose a dual control one-stage framework that decouples the reference image into two levels for flexible control: High-level identity information and low-level texture information, where the identity information figures out the shape of the face and the texture information depicts the component-aware texture. To synthesize high-quality results, we design two novel modules referred to as Half-AdaIN and Component-Wise Style Injector (CWSI) to inject the two kinds of control information into the inpainting processing. Our method produces realistic results with identity and texture control faithful to reference images. To the best of our knowledge, it is the first work to concurrently apply identity and component-level controls in face inpainting to promise more precise and controllable results.

Federated learning has become an emerging hot research field in industry because of its ability to perform large-scale distributed learning while preserving data privacy. However, recent studies have shown that in the actual use of federated learning, there are device heterogeneity and data not identically and independently distributed (Non-IID) characteristics between client nodes, which will affect the effect of federated learning. In this work, we propose resilient reinforcement federated learning (R$^{2}$Fed), a R$^{2}$Fed method, which applies reinforcement learning to federated learning and uses reinforcement learning for weighted fusion of client models instead of average fusion. We conduct experiments on object detection, object classification, and sentiment classification tasks in the context of Non-IID and heterogeneity, and the experimental results show that the R$^{2}$Fed method outperforms traditional federated learning, increasing the average accuracy by 4.7%. Experiments also demonstrate that R$^{2}$Fed is resilient to federation attacks.

In this article, we propose a comprehensive approach to a responsible artificial intelligence (AI)-based software architecture for the digitalization of industry drawings, serving as a software engineering reference for responsible AI in other industry domains.

Since the data samples on client devices are usually non-independent and non-identically distributed (non-IID), this will challenge the convergence of federated learning (FL) and reduce communication efficiency. This paper proposes FedQMIX, a node selection algorithm based on multi-agent reinforcement learning(MARL), to address these challenges. Firstly, we observe a connection between model weights and data distribution, and a clustering algorithm can group clients with similar data distribution into the same cluster. Secondly, we propose a QMIX-based mechanism that learns to select devices from clustering results in each communication round to maximize the reward, penalizing the use of more communication rounds and thereby improving the communication efficiency of FL. Finally, experiments show that FedQMIX can reduce the number of communication rounds by 11% and 30% on the MNIST and CIFAR-10 datasets, respectively, compared to the baseline algorithm (Favor).

With the convergence of modern computing technology and social sciences, both theoretical research and practical applications of social computing have been extended to new domains. In particular, social computing was significantly influenced by the recent advances of artificial intelligence (AI). However, the conventional technologies of AI have various drawbacks in dealing with complicated and dynamic problems. Such deficiency can be rectified by hybrid human-artificial intelligence (H-AI), which integrates both human intelligence and AI into one unity, forming a new enhanced intelligence. H-AI in dealing with social problems shows some advantages over the conventional AI. This article firstly reviews the latest research progresses of AI in social computing. Secondly, it summarizes typical challenges AI faces in social computing, which motivate the necessity to introduce H-AI to tackle social-oriented problems. Finally, we discuss the concept of H-AI and propose a holistic architecture of H-AI in social computing, which consists of three layers: object layer, intelligent processing layer, and application layer. The proposed architecture shows that H-AI has significant advantages over AI in solving social problems.

To improve text sentiments classification issues, such as information loss and insensitivity to spatial information, this paper proposes a text sentiment classification model based on the capsule network (T-Caps), which uses the Transformer to extract low-level text features. The method iteratively updates capsule network parameters through optimized dynamic routing algorithms and global parameter sharing, and it obtains the relationship between local features of the text and the overall emotional polarity to save the information integrity of text features. By comparing with multiple models, we find that the Transformer has the strongest feature extraction capability. The experimental results show that our model is capable of extracting more discriminative semantic features and yields a significant performance gain compared to other baseline methods.

Recently, there is a rising interest in perceiving image aesthetics. The existing works deal with image aesthetics as a classification or regression problem. To extend the cognition from rating to reasoning, a deeper understanding of aesthetics should be based on revealing why a high- or low-aesthetic score should be assigned to an image. From such a point of view, the authors propose a model referred to as Neural Aesthetic Image Reviewer, which can not only give an aesthetic score for an image, but also generate a textual description explaining why the image leads to a plausible rating score. Specifically, they propose three models based on shared aesthetically semantic layers and task-specific embedding layers at a high level for performance improvement on different tasks. To facilitate researches on this problem, they collect the AVA-Reviews dataset, which contains 52,118 images and 312,708 comments in total. Through multi-task learning, the proposed models can rate aesthetic images as well as produce comments in an end-to-end manner. It is confirmed that the proposed models outperform the baselines according to the performance evaluation on the AVA-Reviews dataset. Moreover, they demonstrate experimentally that the authors’ model can generate textual reviews related to aesthetics, which are consistent with human perception.

Pore pressure plays a crucial role in various fields, including geology, petroleum engineering, hydrology, and carbon capture, utilization, and storage (CCUS). Traditional pore pressure prediction methods, which rely on physical models, often fail to meet the required accuracy standards due to their inflexibility and technical parameterization. Although machine learning methods have been increasingly used for pore pressure prediction, they struggle to integrate domain-specific knowledge and capture the inherent temporal dynamics of the data, resulting in suboptimal prediction accuracy. To address these limitations, this study introduces the Knowledge-Aware TFT model, which determines the key parameter attributes for pore pressure prediction by comprehensively analyzing well logging data, rock mechanics theory, and pore pressure evolution theory. Additionally, the model effectively captures the temporal dynamic characteristics of the data, thereby enhancing the accuracy of pore pressure prediction. To validate the model's accuracy, uncertainty theory is employed to evaluate the experimental results, providing a description of the uncertainty associated with pore pressure prediction and supporting credibility assessment. A case study involving three vertical wells in Block X, China, compares the Knowledge-Aware TFT model with SVM, XGBoost, MLP, and LSTM models, further confirming its superior prediction accuracy (Rˆ2=0.997, RMSE=0.135) and demonstrating its practicality in pore pressure prediction. In conclusion, the proposed pore pressure prediction model not only improves prediction accuracy but also achieves excellent results in practical applications, making it a viable alternative to traditional technologies.

Background: Sphingomonas is a kind of microbial resources used for biodegradation of aromatic compounds. In computational biology, identifying protein coding domains in Sphingomonas genome is known as a challenging problem. Objective: In this work, to address the challenge, we propose a novel method to predict protein coding regions from Sphingomonas genome by 3-base periodicity. Method: In our method, DNA sequences are firstly transformed into wavelet by a so-called 3-base characteristics strategy. After that, sliding windows with certain fixed lengths are developed to identify protein coding regions, in which the initial size of sliding windows and values of thresholds are set by experimentally verified protein data in NCBI library. Results: As results, an experimental verified protein coding domain in congeneric families of Sphingomonas is identified from Sphingomonas genome. Conclusion: This would be with high possibility to encode the similar functioning proteins. As well, some potential protein coding regions are marked by narrowing the forecast areas, and then an extensible sliding window strategy is used to improve predictive accuracy.

Video data has become the largest source of big data. Owing to video data's complexities, velocity, and volume, public security and other surveillance applications require efficient, intelligent runtime video processing. To address these challenges, a proposed framework combines two cloud-computing technologies: Storm stream processing and Hadoop batch processing. It uses deep learning to realize deep intelligence that can help reveal knowledge hidden in video data. An implementation of this framework combines five architecture styles: service-oriented architecture, publish-subscribe, the Shared Data pattern, MapReduce, and a layered architecture. Evaluations of performance, scalability, and fault tolerance showed the framework's effectiveness. This article is part of a special issue on Software Engineering for Big Data Systems.

There is a trend that different computing paradigms such as cloud computing, pervasive, and mobile computing are converging with each other. Due to this convergence, there arise unprecedented complexities, such as huge number of computing devices, flexibilities and adaptation of service infrastructure (infrastructure elasticity) for fitting dynamics of large smart city applications, expectations of powerful computing and storage capabilities on handhold devices, and so on. Therefore, a supporting infrastructure is needed, which can flexibly switch services at run time and can be used to enhance capabilities for small devices through component/service migrations. In this paper, we propose an elastic open service gateway initiative (OSGi)-based pervasive cloud (OSGi-PC) infrastructure which can make use of both the cloud computing capabilities and the component flexibilities from OSGi. OSGi-PC provides flexible management of component migrations between small devices themselves and powerful nodes in between, which is remaining a critical challenge for enabling mobile clouds. We have evaluated the OSGi-PC in terms of performance for adaptive service provision, and power consumption during service adaptation, performance and power consumption for component migrations in different scenarios, which show the usability of OSGi-PC.