Developing robust Light Field Image Quality Assessment (LFIQA) methods is essential for advancing light field imaging technologies and enhancing immersive visual experiences. However, existing blind LFIQA methods often lack accuracy and performance, failing to adequately capture the multi-dimensional aspects of light field image (LFI) quality. To address these limitations, we propose a novel blind LFIQA model that incorporates a Fourier Channel Attention Block (FCAB) and multi-scale feature fusion. The FCAB enhances feature extraction through the Fourier transform in the frequency domain, enabling the capture of more representative quality-related information. Additionally, the model incorporates multi-scale feature maps with varying receptive fields through a progressive fusion strategy, enabling a comprehensive evaluation of diverse quality dimensions. To further improve regression accuracy, we employ Kolmogorov-Arnold Networks (KANs), which outperform traditional Multi-Layer Perceptrons (MLPs). Extensive experimental results show that the proposed model surpasses existing LFIQA methods.

Recent work has shown that scene text recognition (STR) models are vulnerable to adversarial examples. Different from non-sequential vision tasks, the output sequence of STR models contains rich information. However, existing adversarial attacks against STR models can only lead to a few incorrect characters in the predicted text. These attack results still carry partial information about the original prediction and could be easily corrected by an external dictionary or a language model. Therefore, we propose the Multi-Population Coevolution Search (MPCS) method to attack each character in the image. We first decompose the global optimization objective into sub-objectives to solve the attack pixel concentration problem existing in previous attack methods. While this distributed optimization paradigm brings a new joint perturbation shift problem, we propose a novel coevolution energy function to solve it. Experiments on recent STR models show the superiority of our method. The code is available at \url{https://github.com/Lee-Jingyu/MPCS}.

Digital Twin technology, initially conceptualized during the NASA's Apollo program, has evolved into a transformative tool for system health management, particularly in aviation. By integrating high-fidelity simulations, real-time sensor data, and predictive analytics, DTs enable significant innovation in Prognostics and Health Management methods. This paper explores the application of DTs in health management for complex aircraft systems, focusing on two critical subsystems: Flight Control Electrical Actuators and Main Landing Gear. Leveraging MATLAB Simscape, modular DT frameworks were developed to simulate these systems under nominal, degraded, and fault conditions. The inclusion of fault injection models enables the generation of realistic datasets to support predictive maintenance, alleviating difficulties in data availability. Two case studies are presented to illustrate the potential of DT-based approaches to reduce downtime, optimizing performance, and enhancing system reliability. This paper provides a comparative analysis of existing DT tools, highlighting their capabilities and limitations in aerospace contexts. While platforms such as MATLAB Simulink and ANSYS Twin Builder offer robust modeling capabilities, operational tools like AVIATAR and IBM Maximo excel in fleet management and predictive analytics. This comparison highlights the need for tailored DT solutions that balance real-time capabilities, scalability, and configurability. This study contributes to the growing body of knowledge on DT technology, offering insights into its role in enhancing aviation safety, efficiency, and sustainability. It serves as a guide for applying DT-based health management, paving the way for broader adoption in next-generation aerospace systems.

Abstract Point cloud normal estimation underpins many 3D vision and graphics applications. Precise normal estimation in regions of sharp curvature and high‐frequency variation remains a major bottleneck; existing learning‐based methods still struggle to isolate fine geometry details under noise and uneven sampling. We present FAHNet, a novel frequency‐aware hierarchical network that precisely tackles those challenges. Our Frequency‐Aware Hierarchical Geometry (FAHG) feature extraction module selectively amplifies and merges cross‐scale cues, ensuring that both fine‐grained local features and sharp structures are faithfully represented. Crucially, a dedicated Frequency‐Aware geometry enhancement (FA) branch intensifies sensitivity to abrupt normal transitions and sharp features, preventing the common over‐smoothing limitation. Extensive experiments on synthetic benchmarks (PCPNet, FamousShape) and real‐world scans (SceneNN) demonstrate that FAHNet outperforms state‐of‐the‐art approaches in normal estimation accuracy. Ablation studies further quantify the contribution of each component, and downstream surface reconstruction results validate the practical impact of our design.

Text recognition in textures is a pressing need for effective contextual perception in our daily life. In OCR scope, deep learning methods can meet the needs of real-time scenarios. However, characters with different textures, such as printed and handwritten characters leads OCR to deal with different scenarios. To this, in this paper, we focus on the implementation and optimization of text recognition using deep learning methods. In particular, we go with a two-stage OCR approach based on deep learning methods to detect and recognize Traditional Chinese characters including handwritten and printed while improving the accuracy. In particular, DBNet [1] is used in the text detection stage and CRNN [2] (baseline) and ABINet [3] (advanced) models in the text recognition stage. For the ABINet, we modify fusion module by using attention mechanism. For the CRNN, we utillize combined loss function. The experiment results show that The CRNN model with the combined loss function improves nearly 4.2% compared with the baseline CRNN. The ABINet model with new fusion module achieves 92.30% on the Traditional Chinese recognition dataset, improves nearly 0.8% compared to the original ABINet. Notably, we also have our own Traditional Chinese handwritten datasets for text detection and recognition. In fine, we believe that our endeavour can help grow the community in a better way.

The recommendation task with a textual corpus aims to model customer preferences from both user feedback and item textual descriptions. It is highly desirable to explore a very deep neural network to capture the complicated nonlinear preferences. However, training a deeper recommender is not as effortless as simply adding layers. A deeper recommender suffers from the gradient vanishing/exploding issue and cannot be easily trained by gradient-based methods. Moreover, textual descriptions probably contain noisy word sequences. Directly extracting feature vectors from them can harm the recommender’s performance. To overcome these difficulties, we propose a new recommendation method named the HighwAy recoMmender (HAM). HAM explores a highway mechanism to make gradient-based training methods stable. A multi-head attention mechanism is devised to automatically denoise textual information. Moreover, a block coordinate descent method is devised to train a deep neural recommender. Empirical studies show that the proposed method outperforms state-of-the-art methods significantly in terms of accuracy.

This article presents a 14-bit 200-Ms/s pipelined analog-to-digital converter (ADC) for wide input frequency range. The ADC adopts a sample-and-hold amplifier-less (SHA-less) 3.5-bit front-end stage. A dedicated path combining architecture is proposed for the flash sub-ADC-based pipeline ADC to reduce the aperture error, and expand the ADC input frequency up to 490 MHz under 200-Ms/s sampling rate. To further improve the accuracy and reduce power, the sampling capacitor splitting and comparator interpolation techniques are used in the first stage. The proposed ADC has been designed and fabricated in a 180-nm CMOS technology with an area of 2.43 mm2including ADC core, the biasing, and the calibration circuit. The measured signal to noise and distortion ratio (SNDR) is 70.6 and 65.6 dB for 30- and 385-MHz inputs, respectively. The SNDR remains 63.1 dB for 490-MHz input. The ADC core consumes 112-mW power under 1.8-V supply. The measured effective resolution bandwidth (ERBW) is about 250 MHz, and the Walden figure of merit (FoM) defined at the ERBW is 116 fJ/conversion-step.

In real-world recommendation tasks, feedback data are usually sparse. Therefore, a recommender’s performance is often determined by how much information that it can extract from textual contents. However, current methods do not make full use of the semantic information. They encode the textual contents either by “bag-of-words” technique or Recurrent Neural Network (RNN). The former neglects the order of words while the latter ignores the fact that textual contents can contain multiple topics. Besides, there exists a dilemma in designing a recommender. On the one hand, we shall use a sophisticated model to exploit every drop of information in item contents; on the other hand, we shall adopt a simple model to prevent itself from over-fitting when facing the sparse feedbacks. To fill the gaps, we propose a recommender named CAMO 1. CAMO employs a multi-layer content encoder for simultaneously capturing the semantic information of multitopic and word order. Moreover, CAMO makes use of adversarial training to prevent the complex encoder from overfitting. Extensive empirical studies show that CAMO outperforms state-of-the-art methods in predicting users’ preferences.

With the recent proliferation of recommendation system, there have been a lot of interests in session-based prediction methods, particularly those based on Recurrent Neural Network (RNN) and their variants. However, existing methods either ignore the dwell time prediction that plays an important role in measuring user's engagement on the content, or fail to process very short or noisy sessions. In this paper, we propose a joint predictor, JUMP, for both user click and dwell time in session-based settings. To map its input into a feature vector, JUMP adopts a novel three-layered RNN structure which includes a fast-slow layer for very short sessions and an attention layer for noisy sessions. Experiments demonstrate that JUMP outperforms state-of-the-art methods in both user click and dwell time prediction.

A 14-bit 250MS/s low power pipeline analog-to-digital converter (ADC) implemented in a 0.18μιη CMOS process is presented in this paper. A SHA-less 3.5-bit front-end is adopted to achieve low power design. An aperture error eliminating technique and flash ADC optimization design techniques such as capacitor splitting, interpolation, and offset calibration, are both used to achieve wideband input even for the front-end with resolution up to 3.5-bit. The post simulation results show this ADC achieves an SNR of 74.6 dB, an SNDR of 74.4 dB and an SFDR of 87.1 dB with a 70MHz input signal, while maintaining an SNR > 72.5 dB and an SFDR > 77.4 dB up to 900MHz input signals. The ADC consumes 120mW from a 1.8 V supply.

This paper presents a low-power 2-channel 12-bit 2GS/S time-interleaved pipelined-SAR ADC in 28nm CMOS process. The design adopts SHA-less front-end, capacitor sharing between stages, current-reused and ping-pong operated MDAC amplifier, and hybrid reference buffer to reduce power consumption and optimize performance. Pre-layout simulation with noise shows that the proposed ADC achieves SNDR and SFDR of 64.1dB and 69.5dB respectively at Nyquist frequency. The effective resolution bandwidth is extended to 3.2GHz. The ADC consumes only 50mW, in which 30mW is dissipated by the wide-band input buffer. At Nyquist frequency, the proposed ADC achieves a Waiden FOM of 19fJ/conv.-step.

As a state-of-the-art recommendation technique, collaborative filtering (CF) methods compute recommendations by leveraging a historical data set of users' ratings for items. So far, the best performing CF methods are latent factor models. Probabilistic matrix factorization (PMF) model, as a widely used latent factor model, offers a probabilistic foundation for regularization. In this paper, we present a novel CF method by incorporating implicit relationship between items into the basic PMF model. Firstly we mine the implicit correlation between items based on a matrix factorization model by utilizing contextual information, and then generalize recommendations by incorporating the obtained item relationship into the basic PMF model. We validate our approach on two datasets, and the experimental results show that the proposed method outperforms several existing CF models.

Data centers are growing in size and complexity driven by trends such as cloud computing and on-line services. Such large data centers pose several challenges for system management. Key among them is anomaly detection which is required to monitor and analyze metrics across several thousands servers and across multiple layers of abstractions to detect anomalous system behavior. In practice, multiple anomaly detection tools are used to continuously raise alarms across multiple metrics and servers. These alarms include both true positives and false alarms. Administrators and management tools act on these alarms for diagnosis and deeper root cause analysis and take appropriate management actions to mitigate the anomalous behaviors. Given the scale and scope of the system, the administrators and management tools are overwhelmed with the large number of alarms at any given instant, many of which are false alarms. It is therefore necessary to prioritize and rank these alarms, so as to take timely actions that maintain the service level agreements for the data center. Existing techniques for such ranking are ad-hoc and not scalable. We propose ranking windows of monitored metrics based on their probability of occurrence. We explain how these probabilities can be computed based either on the false positive rates for which the accompanying anomaly detectors were designed, or, when available, on the probability models underlying the false positive rates. In the simplest case, the ranking procedure reduces to computing the Z-score of the observed measurements and computing a statistic from a window of Z-scores to use as a basis for ranking. The proposed techniques are reliable, lightweight and easy to deploy in the modern data center. We have validated these techniques on synthetic data containing injected anomalies and on data acquired from production data centers.

Online anomaly detection is an important step in data center management, requiring light-weight techniques that provide sufficient accuracy for subsequent diagnosis and management actions. This paper presents statistical techniques based on the Tukey and Relative Entropy statistics, and applies them to data collected from a production environment and to data captured from a testbed for multi-tier web applications running on server class machines. The proposed techniques are lightweight and improve over standard Gaussian assumptions in terms of performance.

The online detection of anomalies is a vital element of operations in data centers and in utility clouds like Amazon EC2. Given ever-increasing data center sizes coupled with the complexities of systems software, applications, and workload patterns, such anomaly detection must operate automatically, at runtime, and without the need for prior knowledge about normal or anomalous behaviors. Further, detection should function for different levels of abstraction like hardware and software, and for the multiple metrics used in cloud computing systems. This paper proposes EbAT - Entropy-based Anomaly Testing - offering novel methods that detect anomalies by analyzing for arbitrary metrics their distributions rather than individual metric thresholds. Entropy is used as a measurement that captures the degree of dispersal or concentration of such distributions, aggregating raw metric data across the cloud stack to form entropy time series. For scalability, such time series can then be combined hierarchically and across multiple cloud subsystems. Experimental results on utility cloud scenarios demonstrate the viability of the approach. EbAT outperforms threshold-based methods with on average 57.4% improvement in accuracy of anomaly detection and also does better by 59.3% on average in false alarm rate with a `near-optimum' threshold-based method.

GlobalWatch is a distributed platform to monitor various resources of grid platforms so as to improve the flexibility and usability of grid systems. In order to enhance the flexibility, we make improvements from three aspects. 1) We take advantage of both centralized and decentralized methods in the architecture and adopt a loosely coupled architecture to get monitored data; 2) We integrate GlobalWatch with VO (Virtual Organization) concept so as to make it compatible with grid circumstances. 3) We design a XML format to make GlobalWatch flexible for users to submit and receive information. In order to enhance the usability, we not only integrate various functions into GlobalWatch client application, but also make the visual effect better than some other monitoring platforms. At last, we conclude from the experimental result that monitored nodes in GlobalWatch are indeed loosely coupled and the transmission of map data is the main bottleneck in server end

When facing the emerging circumstance that the Internet and business globalization have replaced the separation that was typical of the traditional business paradigm, especially the workflow and grid environment, challenges have been made to the traditional static access control model. To solve these challenges, we present a dynamic access control prediction mechanism for service workflow base on Markov chain. We abstract the service workflow in workflow and/or grid to an ordered service sequence and through calculating the executable probability of service in OSS, we create an authorization policy based on the probability value dynamically. A simulation method of the mechanism is presented and practiced in ChinaGrid project