Bias in Large Language Models (LLMs) is increasingly addressed through fairness-oriented techniques. However, in some cases, these approaches may inadvertently remove genuine cultural differences between groups, leading to “over-normalization” or models losing important socio-cultural distinctions. In this work, we introduce OverNormEval, a benchmark designed to detect when an LLM exhibits such over-normalization. We further explore the use of Direct Preference Optimization (DPO) to mitigate over-normalization.

Automated captioning for remote sensing imagery often struggles to balance the high descriptive power of large models with the deployment feasibility of smaller ones. To bridge this gap, this paper introduces ONIR, a LLM-efficient, tag-guided framework that empowers compact language models (1-3B parameters) to achieve state-of-the-art captioning accuracy. Specifically, the proposed approach synthesizes a large-scale pseudo-caption dataset by leveraging GPT-4O on existing segmentation benchmarks. Explicit semantic tags are then extracted to train a multi-label Contrastive Language-Image Pre-Training (CLIP) encoder, providing interpretable visual guidance. To maintain parameter efficiency, the architecture incorporates a simple Multilayer Perceptron (MLP) bridge and a two-stage LoRA fine-tuning strategy. Extensive experiments on standard benchmark dataset, such as UCM and Sydney Captions, demonstrate that ONIR significantly outperforms models up to four times its size (7-13B). By combining superior performance with computational efficiency and tag-based controllability, ONIR offers a highly practical solution for real-world remote sensing applications.

Services based on various forms of video streaming deliver content to clients by compressing originally captured videos using video codecs. During this process, due to limitations such as network bandwidth and display device capabilities, the original resolution is often reduced or a higher compression rate is applied, resulting in smaller data sizes being transmitted. Consequently, the video quality experienced by end users is often degraded. To address this issue, numerous AI-based Video Super-Resolution (VSR) techniques have been proposed. However, in addition to their high architectural complexity, many of these models require a large number of input frames or rely on recurrent frameworks, which further complicate both training and inference. These factors present significant challenges for deployment in real-time video services. In this paper, we propose a lightweight VSR model that achieves high performance even on low-quality, highly compressed video content. We propose an adaptive inference method that dynamically selects between lightweight and enhanced processing modules based on the input video’s resolution, thereby improving the suitability of our approach for real-time video streaming applications. Additionally, we visualize the performance of the proposed model with Grad-CAM to demonstrate its effectiveness compared to existing methods.

The integration of conversational agents into daily life has become increasingly common. However, sustaining deeply engaging and natural interactions remains challenging due to a lack of multimodal datasets capturing personal and emotional nuances. In this paper, we introduce MERCI (Multimodal dataset for Emotionally-aware peRsonalised Conversational In-teractions), a dataset derived from user-robot dialogues involving thirty participants who completed user profile questionnaires covering ten personal topics (e.g., hobbies, music). A conver-sational system called PERCY then engaged with each partici-pant in open-domain conversations, leveraging GPT-4, real-time facial-expression and sentiment analysis to generate contextu-ally appropriate, empathetic responses. MERCI contains 1860 utterances, equating to about 12.5 hours of aligned audio, three-view video, transcripts with timestamps, emotion labels, and sentiment scores. This dataset serves as a reproducible test-bed for tasks such as emotion-aware response generation, multimodal affect recognition, and personalised policy learning. Baseline performance results have been established using advanced models such as BERT, T5, BART, and GPT-3.5/4/4o-mini across gener-ation, regression, and classification. Evaluations through human and automated methods have demonstrated strong naturalness, relevance, and consistency in responses while indicating areas for enhanced personalisation and empathic depth. We expect that MERCI will enhance the development of emotionally intelligent, user-centric conversational AI applications, potentially ranging from social robotics to mental health support.

Recent vision-language models (e.g., CLIP) have demonstrated remarkable class-generalizable ability to unseen classes in few-shot anomaly segmentation (FSAS), leveraging supervised prompt learning or fine-tuning on seen classes. However, their cross-category generalization largely depends on prior knowledge of real seen anomaly samples. In this paper, we propose a novel framework, namely DictAS, which enables a unified model to detect visual anomalies in unseen object categories without any retraining on the target data, only employing a few normal reference images as visual prompts. The insight behind DictAS is to transfer dictionary lookup capabilities to the FSAS task for unseen classes via self-supervised learning, instead of merely memorizing the normal and abnormal feature patterns from the training set. Specifically, DictAS mainly consists of three components: (1) Dictionary Construction - to simulate the index and content of a real dictionary using features from normal reference images. (2) Dictionary Lookup - to retrieve queried region features from the dictionary via a sparse lookup strategy. When a query feature cannot be retrieved, it is classified as an anomaly. (3) Query Discrimination Regularization - to enhance anomaly discrimination by making abnormal features harder to retrieve from the dictionary. To achieve this, Contrastive Query Constraint and Text Alignment Constraint are further proposed. Extensive experiments on seven public industrial and medical datasets demonstrate that DictAS consistently outperforms state-of-the-art FSAS methods.

Visual Question Answering (VQA) in remote sensing (RS) is pivotal for interpreting Earth observation data. However, existing RS VQA datasets are constrained by limitations in annotation richness, question diversity, and the assessment of specific reasoning capabilities. This paper introduces Remote Sensing Vision Language Model Question Answering (RSVLM-QA) dataset, a new large-scale, content-rich VQA dataset for the RS domain. RSVLM-QA is constructed by integrating data from several prominent RS segmentation and detection datasets: WHU, LoveDA, INRIA, and iSAID. We employ an innovative dual-track annotation generation pipeline. Firstly, we leverage Large Language Models (LLMs), specifically GPT-4.1, with meticulously designed prompts to automatically generate a suite of detailed annotations including image captions, spatial relations, and semantic tags, alongside complex caption-based VQA pairs. Secondly, to address the challenging task of object counting in RS imagery, we have developed a specialized automated process that extracts object counts directly from the original segmentation data; GPT-4.1 then formulates natural language answers from these counts, which are paired with preset question templates to create counting QA pairs. RSVLM-QA comprises 13,820 images and 162,373 VQA pairs, featuring extensive annotations and diverse question types. We provide a detailed statistical analysis of the dataset and a comparison with existing RS VQA benchmarks, highlighting the superior depth and breadth of RSVLM-QA's annotations. Furthermore, we conduct benchmark experiments on Six mainstream Vision Language Models (VLMs), demonstrating that RSVLM-QA effectively evaluates and challenges the understanding and reasoning abilities of current VLMs in the RS domain. We believe RSVLM-QA will serve as a pivotal resource for the RS VQA and VLM research communities, poised to catalyze advancements in the field. The dataset, generation code, and benchmark models are publicly available at https://github.com/StarZi0213/RSVLM-QA.

Neural Radiance Fields (NeRFs) have shown great potential in modeling 3D scenes. Dynamic NeRFs extend this model by capturing time-varying elements, typically using deformation fields. The existing dynamic NeRFs employ a similar Eulerian representation for both light radiance and deformation fields. This leads to a close coupling of appearance and motion and lacks a physical interpretation. In this work, we propose Dynamic Appearance Particle Neural Radiance Field (DAP-NeRF), which introduces particle-based representation to model the motions of visual elements in a dynamic 3D scene. DAP-NeRF consists of the superposition of a static field and a dynamic field. The dynamic field is quantized as a collection of appearance particles, which carries the visual information of a small dynamic element in the scene and is equipped with a motion model. All components, including the static field, the visual features and the motion models of particles, are learned from monocular videos without any prior geometric knowledge of the scene. We develop an efficient computational framework for the particle-based model. We also construct a new dataset to evaluate motion modeling. Experimental results show that DAP-NeRF is an effective technique to capture not only the appearance but also the physically meaningful motions in a 3D dynamic scene. Code is available at:https://github.com/Cenbylin/DAP-NeRF.

Influence Maximization is the task of selecting optimal nodes maximising the influence spread in social networks. This study proposes a Discretized Quantum-based Salp Swarm Algorithm (DQSSA) for optimizing influence diffusion in social networks. By discretizing meta-heuristic algorithms and infusing them with quantum-inspired enhancements, we address issues like premature convergence and low efficacy. The proposed method, guided by quantum principles, offers a promising solution for Influence Maximisation. Experiments on four real-world datasets reveal DQSSA's superior performance as compared to established cutting-edge algorithms.

In this paper we propose a new version of penalty-based aggregation functions, the Multi Cost Aggregation choosing functions (MCAs), in which the function to minimize is constructed using a convex combination of two relaxed versions of restricted equivalence and dissimilarity functions instead of a penalty function. We additionally suggest two different alternatives to train a MCA in a supervised classification task in order to adapt the aggregation to each vector of inputs. We apply the proposed MCA in a Motor Imagery-based Brain Computer Interface (MI-BCI) system to improve its decision making phase. We also evaluate the classical aggregation with our new aggregation procedure in two publicly available datasets. We obtain an accuracy of 82.31% for a left vs. right hand in the Clinical BCI challenge (CBCIC) dataset, and a performance of 62.43% for the four-class case in the BCI Competition IV 2a dataset compared to a 82.15% and 60.56% using the arithmetic mean. Finally, we have also tested the goodness of our proposal against other MI-BCI systems, obtaining better results than those using other decision making schemes and Deep Learning on the same datasets.

Recent works on fake news detection have shown the efficacy of using emotions as a feature for improved performance. However, the cross-domain impact of emotion-guided features for fake news detection still remains an open problem. In this work, we propose an emotion-guided, domain-adaptive, multi-task approach for cross-domain fake news detection, proving the efficacy of emotion-guided models in cross-domain settings for various datasets.

Social networks have enabled user-specific advertisements and recommendations on their platforms, which puts a significant focus on Influence Maximisation (IM) for target advertising and related tasks. The aim is to identify nodes in the network which can maximize the spread of information through a diffusion cascade. We propose a community structures-based approach that employs K-Shell algorithm with community structures to generate a score for the connections between seed nodes and communities. Further, our approach employs entropy within communities to ensure the proper spread of information within the communities. We validate our approach on four publicly available networks and show its superiority to four state-of-the-art approaches while still being relatively efficient.

Abstract Mental task classification (MTC), based on the electroencephalography (EEG) signals is a demanding brain–computer interface (BCI). It is independent of all types of muscular activity. MTC-based BCI systems are capable to identify cognitive activity of human. The success of BCI system depends upon the efficient feature representation from raw EEG signals for classification of mental activities. This paper mainly presents on a novel feature representation (formation of most informative features) of the EEG signal for the both, binary as well as multi MTC, using a combination of some statistical, uncertainty and memory- based coefficient. In this work, the feature formation is carried out in the two stages. In the first stage, the signal is split into different oscillatory functions with the help of three well-known empirical mode decomposition (EMD) algorithms, and a new set of eight parameters (features) are calculated from the oscillatory function in the second stage of feature vector construction. Support vector machine (SVM) is used to classify the feature vectors obtained corresponding to the different mental tasks. This study consists the problem formulation of two variants of MTC; two-class and multi-class MTC. The suggested scheme outperforms the existing work for the both types of mental tasks classification.

Micro electro mechanical system (MEMS) tunable capacitors (TC) are major elements in ac voltage reference sources (VRS). Physical parametric uncertainties, external electrostatic disturbance, and measurement noise malfunction their operation and ruin the preciseness of the VRS output voltage. Our objective problem is the design of a controller to cope with the mentioned parametric uncertainties, noise, and disturbance. Our applied method is the design and employment of a proportional integral fuzzy adaptive sliding mode controller (FASMC). Both terms of matched and unmatched uncertainties as well as external disturbance and measurement noise are all addressed in this article to generate a stable VRS output for the first time. Not only does this article contribute to the employment of a FASMC to enhance robustness in the drive of the capacitor, but also it benefits the reduction of the chattering effect due to fuzzy regulation in the switching term of the control law. Moreover, the automatic fuzzy adjustment in the controller, which is used for estimation of the coefficients in the sliding surface error dynamical equation, facilitates the specification of those coefficients, which can be time consuming in simulation affairs. Clarifying the importance of the proposed fuzzy adaptive sliding mode controller, this article reviews some previous controllers for MEMS TC in comparison with the proposed fuzzy controller, demonstrating its enhancement in comparison with previous schemes.

To handle the fuzziness and spatial uncertainties among pixels entailed in color images, this paper proposes a novel fuzzy entropy function for multi-threshold image segmentation based on the energy curve concept and minimum fuzzy entropy criterion. The proposed energy curve based new fuzzy entropy function (ECFE) considers intensity distribution and spatial contextual information among the pixels. To improve efficiency and threshold selection process of the method, cuckoo search algorithm is employed. For comparison, backtracking search algorithm, and Lévy flight based firefly algorithm included. Comparison with recent color image multilevel segmentation techniques presented to test the effectiveness of the proposed algorithm. The performance of the proposed technique is evaluated using different satellite and natural color images. Quantitative and qualitative results demonstrate that the proposed algorithm is highly accurate, robust, and efficient for color image multilevel segmentation.

Alzheimer's Disease (AD) is one of the most common forms of neuropsychological disorder in elderly people. It is a slow progressive disease affecting the brain cells. This affects the cognitive abilities of people and their daily activities. During the course of the disease, memory gets brutally affected too. Working as well as long-term declarative memory deteriorates in AD patients. Due to this deterioration of the memory, AD patients tend to show a decline in their communicative skills as well. This decline is reflected in their speech. AD patients usually have poor grammar along with very low coherent ideas. Also, they tend to repeat the words very often and hence become unclear on the message they are trying to convey. As the disease progresses, the speech is completely impaired, and the patients are left to sing or utter words that are totally out of context. Stopwords are the words that are most commonly used in language and it is often hypothesized that AD patients use them much often as compared to Control Normal (CN) subjects. It is seen that due to the degeneration of brain cells in AD patients, they have a tendency to use a lot of stopwords to fill their perplexities in their statements. In this paper, the usefulness of the stopwords in capturing the linguistic information of the patients suffering from AD are discussed. Learning algorithms are evaluated by including stopwords and dropping stopwords at preprocessing to draw comparisons.

Zero-shot learning (ZSL) aims to recognize instances belonging to unseen categories which are not available at training time. Previous ZSL models learn a projection function from the visual feature space to a semantic space which contains a description of the categories. The semantic attributes are often correlated with each other at the semantic space and it is not appropriate to learn them independently. Existing ZSL methods are designed to work on complete descriptions of the semantic attributes. However, because these attributes are human-designed values, they might be incomplete or contains noisy values which may affect the recognition performance of many existing ZSL models. This paper proposes a novel zero-shot learning approach (ZSL-MSA) to handle missing and noisy semantic attributes during the training process. Significantly, the proposed method learns a supplementary attribute matrix by exploiting the attribute correlation. The proposed method also learns the relevant feature coefficients in the projection matrix to identify the correlated attribute space. Th proposed method also adopts l1regularization norm to select the relevant sparse features. A constrained optimization function is formulated and solved using the accelerated proximal gradient method. Extensive experiments on three benchmark datasets using ZSL and generalized ZSL demonstrate the effectiveness of the proposed method.

With the recent surge in threats to public safety, the security focus of several organizations has been moved towards enhanced intelligent screening systems. Conventional X-ray screening, which relies on the human operator is the best use of this technology, allowing for the more accurate identification of potential threats. This paper explores X-ray security imagery by introducing a novel approach that generates realistic synthesized data, which opens up the possibility of using different settings to simulate occlusion, radiopacity, varying textures, and distractors to generate cluttered scenes. The generated synthetic data is effective in the training of deep networks. It allows better generalization on training data to deal with domain adaptation in the real world. The extensive set of experiments in this paper provides evidence for the efficacy of synthetic datasets over human-annotated datasets for automated X-ray security screening. The proposed approach outperforms the state-of-the-art approach for a diverse threat object dataset on mean Average Precision (mAP) of region-based detectors and classification/regression-based detectors.

Word representation has always been an important research area in the history of natural language processing (NLP). Understanding such complex text data is imperative, given that it is rich in information and can be used widely across various applications. In this survey, we explore different word representation models and its power of expression, from the classical to modern-day state-of-the-art word representation language models (LMS). We describe a variety of text representation methods, and model designs have blossomed in the context of NLP, including SOTA LMs. These models can transform large volumes of text into effective vector representations capturing the same semantic information. Further, such representations can be utilized by various machine learning (ML) algorithms for a variety of NLP-related tasks. In the end, this survey briefly discusses the commonly used ML- and DL-based classifiers, evaluation metrics, and the applications of these word embeddings in different NLP tasks.

Computer vision-based vehicle detection techniques are widely used in real-world applications. However, most of these techniques aim to detect only single-view vehicles, and their performances are easily affected by partial occlusion. Therefore, this paper proposes a novel multi-view vehicle detection system that uses a part model to address the partial occlusion problem and the high variance between all types of vehicles. There are three features in this paper; firstly, different from Deformable Part Model, the construction of part models in this paper is visual and can be replaced at any time. Secondly, this paper proposes some new part models for detection of vehicles according to the appearance analysis of a large number of modern vehicles by the active learning algorithm. Finally, this paper proposes the method that contains color transformation along with the Bayesian rule to filter out the background to accelerate the detection time and increase accuracy. The proposed method outperforms other methods on given dataset.

In this paper, classification of mental task-root brain-computer interfaces (BCIs) is being investigated. The mental tasks are dominant area of investigations in BCI, which utmost interest as these system can be augmented life of people having severe disabilities. The performance of BCI model primarily depends on the construction of features from brain, electroencephalography (EEG), signal, and the size of feature vector, which are obtained through multiple channels. The availability of training samples to features are minimal for mental task classification. The feature selection is used to increase the ratio for the mental task classification by getting rid of irrelevant and superfluous features. This paper suggests an approach to augment the performance of a learning algorithm for the mental task classification on the utility of power spectral density (PSD) using feature selection. This paper also deals a comparative analysis of multivariate and univariate feature selection for mental task classification. After applying the above stated method, the findings demonstrate substantial improvements in the performance of learning model for mental task classification. Moreover, the efficacy of the proposed approach is endorsed by carrying out a robust ranking algorithm and Friedman's statistical test for finding the best combinations and compare various combinations of PSD and feature selection methods.

The Internet has resulted in cyber-threats and cyber-crimes, which can occur anywhere at any time. Among various cyber threats, modern malware with applied metamorphosis and polymorphic technology is a concern as it can proliferate to advanced variants from its original shape. The typical malware analysis methods, including signature-based approach, remain vulnerable to such advanced variants. This paper proposes a visualization-based approach for malware analysis using the state-of-the-art Convolution Neural Network (CNN) model such as ResNeXt, which had achieved outstanding performance in image classifications with competitive computational complexity. The proposed method transforms the attributes of raw malware binary executable files to greyscale images for further analysis by well-established deep learning models. The greyscale images, which result of data transformation for visualization, are classified using ResNeXt. The experiment results show that the proposed solution achieves 98.32% and 98.86% of accuracy in malware classification on Malimg dataset and modified Malimg dataset, respectively. The proposed method outperforms other comparable methods in terms of classification accuracy and requires similar level of computational power.

Text in natural scene image portrays rich semantic information that plays an important role in content analysis. However, apart from Arabic text in documents, the text in natural scene images exhibit much higher diversity and variability, especially in uncontrolled circumstances. In this paper, a hybrid feature extraction approach is presented to detect extremal region of Arabic scene text. The binary image and image mask are considered as a variant of input image and look for concurrent extremal regions in both images. After determination of conjoined extremal points, the scale invariant technique is applied to consider those invariant points which are common in both images based on their coordinate positions. To evaluate the performance, a multidimensional long short term memory (LSTM) network is adapted and obtained 94.21% accuracy for word recognition on unconstrained Arabic scene text recognition (ASTR) dataset.

Right from the very beginning, the text has vital importance in human life. As compared to the vision-based applications, preference is always given to the precise and productive information embodied in the text. Considering the importance of text, recognition, and detection of text is also equally important in human life. This paper presents a deep analysis of recent development on scene text and compare their performance and bring into light the real modern applications. Future potential directions of scene text detection and recognition are also discussed.

Text mining in the biomedical domain is increasingly important as the volume of biomedical documents increases. Thanks to advances in natural language processing (NLP), extracting valuable information from the biomedical literature is gaining popularity among researchers, and deep learning has enabled the development of effective biomedical text mining models. However, directly applying advancements in NLP to biomedical sources often yields unsatisfactory results, due to a word distribution drift from the general language domain corpora to specific biomedical corpora, and this drift introduces linguistic ambiguities. To overcome these challenges, this paper presents a novel method for biomedical named entity-recognition (BioNER) through hierarchically fusing representations from BioBERT, which is trained on biomedical corpora and Deep contextual-level word embeddings to handle the linguistic challenges within biomedical literature. Proposed text representation is then fed to attention-based Bi-directional Long Short Term Memory (BiLSTM) with Conditional random field (CRF) for the BioNER task. The experimental analysis shows that our proposed end-to-end methodology outperforms existing state-of-the-art methods for the BioNER task.

Alzheimer's disease (AD) is a neurodegenerative disorder resulting in memory loss and cognitive decline caused due to the death of brain cells. It is the most common form of dementia and accounts for 60-80% of all dementia cases. There is no single test for diagnosis of AD, the doctors rely on medical history, neuropsychological assessments, computed tomography (CT) or magnetic resonance imaging (MRI) scan of the brain, etc. to confirm a diagnosis. In terms of the treatment, currently, there is neither a cure nor any way to slow the progression of AD. However, for people with mild or moderate stages of this disease, there are some medications available to temporarily reduce symptoms and help to improve quality of life. Hence, early diagnosis of AD is extremely crucial for overall better management of the disease. The researches have shown some relation between neuropsychological scores and atrophies of the brain. This can be leveraged for the early diagnosis of AD. This paper makes use of feature selection techniques to extract the most important features in the diagnosis of AD. This paper demonstrates the need to combine neuropsychological scores like mini-mental state examination (MMSE) with MRI features to provide better decisional space for early diagnosis of AD. Through the experiments, including MMSE along with other features are found to improve the classification of AD, significantly.

Alzheimer’s disease is an incurable neurodegenerative disease primarily affecting the elderly population. Efficient automated techniques are needed for early diagnosis of Alzheimer’s. Many novel approaches are proposed by researchers for classification of Alzheimer’s disease. However, to develop more efficient learning techniques, better understanding of the work done on Alzheimer’s is needed. Here, we provide a review on 165 papers from 2005 to 2019, using various feature extraction and machine learning techniques. The machine learning techniques are surveyed under three main categories: support vector machine (SVM), artificial neural network (ANN), and deep learning (DL) and ensemble methods. We present a detailed review on these three approaches for Alzheimer’s with possible future directions.

It is believed that failures of people's reaction to emergencies occurred during driving are closely related to the inhibitory mechanism of brain's operations. To investigate the role of this function in emergency driving, two virtual realistic driving conditions based on stop signal task were designed and time limitation was manipulated to increase the stress in one condition. Sixteen subjects with behavioral encephalography recordings were collected and analyzed. By comparing successful and unsuccessful stop trials with event-related spectral perturbation analysis, δ and θ band power increases in frontal and central areas are correlated with driving inhibitory control of the brain. Moreover, β and y band power in frontal and central areas showed more increases upon stress condition. Time pressure in driving could adjust the operation of brain's inhibition control, to benefit the people's reactive ability upon emergency.

The ubiquitous availability of smartphones and the increasing popularity of social media provide a platform for users to express their feelings, including suicidal ideation. Suicide prevention by suicidal ideation detection on social media lights the path to controlling the rapidly increasing suicide rates amongst youth. This paper proposes a diverse set of features and investigates into feature selection using the Firefly algorithm to build an efficient and robust supervised approach to classifying tweets with suicidal ideation. The development of a suicidal language to create three diverse, manually annotated datasets leads to the validation of the proposed model. An in-depth result and error analysis lead to an accurate system for monitoring suicidal ideation on social media along with the discovery of optimal feature subsets and selection methods using a penalty based Firefly algorithm.

Prior studies have proposed methods to recover multi-channel electroencephalography (EEG) signal ensembles from their partially sampled entries. These methods depend on spatial scenarios, yet few approaches aiming to a temporal reconstruction with lower loss. The goal of this study is to retrieve the temporal EEG signals independently which was overlooked in data pre-processing. We considered EEG signals are impinging on tensor-based approach, named nonlinear Canonical Polyadic Decomposition (CPD). In this study, we collected EEG signals during a resting-state task. Then, we defined that the source signals are original EEG signals and the generated tensor is perturbed by Gaussian noise with a signal-to-noise ratio of 0 dB. The sources are separated using a basic nonnegative CPD and the relative errors on the estimates of the factor matrices. Comparing the similarities between the source signals and their recovered versions, the results showed significantly high correlation over 95%. Our findings reveal the possibility of recoverable temporal signals in EEG applications.

Electroencephalogram signal is the signal used for the detection of a neurological disorder as epilepsy disorder, sleep disorder and many more. The types of EEG signal gives the hidden information regarding the distribution of the data that may consist of a large volume of the poor and noisy signal. In order to reduce the outlier effects and noise, incorporation of prior knowledge in the model, universum may help and enhance the better generalization ability of the model. This paper proposes a regularized universum twin support vector machine (RUTWSVM) for classification of the healthy and seizure EEG signals. Here, the selection of the universum data points is obtained in two ways (i). Universum data has been generated from the healthy and seizure EEG signals itself and (ii). Interictal EEG signal has been used as universum data which may help to handle the outlier effects. Further, various feature selection techniques are applied to extract the important noise free features from the EEG signals. We have performed a comparative analysis of proposed RUTWSVM with USVM and UTWSVM to classify the EEG signals as well as benchmark real-world datasets in an optimum way. The experiment results clearly exhibit the applicability and usability of the proposed RUTWSVM with interictal EEG signals as universum data points as well as benchmark real-world datasets.

This paper proposes applying recurrent residual network (RRN) for analyzing electroencephalogram (EEG) data captured during a simulated sustained attention driving task. We first address the suitableness of utilizing residual structure as well as adopting recurrent structure for EEG signal processing. Then based on these descriptions a recurrent residual network is tailored and depicted in detail. Thirdly we use an EEG dataset obtained from a sustained-attention experiment for our model justification. By applying the RRN model to the experimental data and via the competitive result achieved, we demonstrate the elegance of the proposed model. At last, we discuss the characteristics of the learned filters and their interpretations from EEG frequency band perspectives.

Retinopathy of Prematurity (ROP) is a disease which requires immediate precautionary measures to prevent blindness in the infants, and this condition is prevalent in premature babies in all the underdeveloped, developing, and in the developed countries as well. This paper proposes a tool by which the stage and zones of Retinopathy of Prematurity in infants can be diagnosed easily. This tool takes the input from the Retcam and detects the stage, zone, and gives a rating of 1 to 9 for classifying the severity of the disease in the infants. This is achieved by extracting the optic disc, marking the ridge, and the distance of the optic nerve. This tool can be easily used by nurses and paramedics, unlike the existing technologies which require the guidance of a specialist to come to a conclusion.

Observed data often have multiple labels with respect to different aspects. For example, a picture can have one label specifying the contents in terms of the object category such as aeroplane, building, cat, etc. and in the meanwhile have another label describing the image style such as photo-realistic or artistic. The central idea of this work is that any annotation of the data contains precious knowledge and is not to be foregone: an analytic task focusing on one aspect of the data can benefit from the knowledge transferred from the other aspects. We propose a passive knowledge transfer scheme for deep neural network training based on the generative adversarial nets (GANs). The adversarial training scheme encourages the nets to encode data into representations that are both discriminative for the target aspect and invariant with respect to the irrelevant aspects. We show that the scheme mixes the conditional distributions of the encoded data on the irrelevant aspects, by the theory on the link between the GAN framework and the Wasserstein metric in distribution spaces. Moreover, we empirically verified the method by i) classifying images despite influence by geometric transform and ii) recognizing the movements (geometric transform) regardless the image contents.

Nowadays, most ofthe vehicle detection methods aim to detect only single-view vehicles, and the performance is easily affected by partial occlusion. Therefore, a novel multi-view vehicle detection system is proposed to solve the problem of partial occlusion. The proposed system is divided into two steps: background filtering and part model. Background filtering step is used to filter out trees, sky and other road background objects. In the part model step, each of the part models is trained by samples collected by using the proposed active learning algorithm. This paper validates the performance of the background filtering method and the part model algorithm in multi-view car detection. The performance of the proposed method outperforms previously proposed methods.

This paper proposes an information completion technique, GAN2C, by imposing dual consistency constraints (2C) to a closed loop encoder-decoder architecture based on the generative adversarial nets (GAN). When adopting deep neural networks as function approximators, GAN2C enables highly effective multi-modality image conversion with sparse observation in the target modes. For empirical demonstration and model evaluation, we show that trained deep neural networks in GAN2C can infer colors for grayscale images, as well as estimate rich 3D information of a scene by densely predicting the depths. The results of the experiments show that in both tasks GAN2C as a generic framework has been comparable to or advanced the state-of-the-art performance which are achieved by highly specialized systems. Code is available at https://github.com/AdalinZhang/GAN2C.

Recently, virtualization in wireless sensor networks (WSNs) has witnessed significant attention due to the growing service domain for Internet of Things (IoT). Related literature on virtualization in WSNs explored resource optimization without considering communication failure in WSNs environments. The failure of a communication link in WSNs impacts many virtual networks running IoT services. In this context, this paper proposes a framework for optimizing fault tolerance (FT) in virtualization in WSNs, focusing on heterogeneous networks for service-oriented IoT applications. An optimization problem is formulated considering FT and communication delay as two conflicting objectives. An adapted nondominated sorting-based genetic algorithm (A-NSGA) is developed to solve the optimization problem. The major components of A-NSGA include chromosome representation, FT and delay computation, crossover and mutation, and nondominance-based sorting. Analytical and simulation-based comparative performance evaluation has been carried out. From the analysis of results, it is evident that the framework effectively optimizes FT for virtualization in WSNs.

The main challenge of attribute reduction in large data applications is to develop a new algorithm to deal with large, noisy, and uncertain large data linking multiple relevant data sources, structured or unstructured. This paper proposes a new and efficient layered-coevolution-based attribute-boosted reduction algorithm (LCQ-ABR*) using adaptive quantum-behavior particle swarm optimization (PSO). First, the quantum rotation angle of an evolutionary particle is updated by a dynamic change of self-adapting step size. Second, a self-adaptive partitioning strategy is employed to group particles into different memeplexes, and the quantum-behavior mechanism with the particles' states depicted by the wave function cooperates to achieve superior performance in their respective memeplexes. Third, a new layered coevolutionary model with multiagent interaction is constructed to decompose a complex attribute set, and it can self-adapt the attribute sizes among different layers and produce the reasonable decompositions by exploiting any interdependence among multiple relevant attribute subsets. Fourth, the decomposed attribute subsets are evolved to compute the positive region and discernibility matrix by using their best quantum particles, and the global optimal reduction set is induced successfully. Finally, extensive comparative experiments are provided to illustrate that LCQ-ABR*has better feasibility and effectiveness of attribute reduction on large-scale and uncertain dataset problems with complex noise as compared with representative algorithms. Moreover, LCQ-ABR*can be successfully applied in the consistent segmentation for neonatal brain three-dimensional MRI, and the consistent segmentation results further demonstrate its stronger applicability.

This study considers the dynamic changes of complexity feature by fuzzy entropy measurement and repetitive steady-state visual evoked potential (SSVEP) stimulus. Since brain complexity reflects the ability of the brain to adapt to changing situations, we suppose such adaptation is closely related to the habituation, a form of learning in which an organism decreases or increases to respond to a stimulus after repeated presentations. By a wearable electroencephalograph (EEG) with Fpz and Oz electrodes, EEG signals were collected from 20 healthy participants in one resting and five-times 15 Hz SSVEP sessions. Moreover, EEG complexity feature was extracted by multi-scale Inherent Fuzzy Entropy (IFE) algorithm, and relative complexity (RC) was defined the difference between resting and SSVEP. Our results showed the enhanced frontal and occipital RC was accompanied with increased stimulus times. Compared with the 1st SSVEP session, the RC was significantly higher than the 5th SSVEP session at frontal and occipital areas (p <; 0.05). It suggested that brain has adapted to changes in stimulus influence, and possibly connected with the habituation. In conclusion, effective evaluation of IFE has a potential EEG signature of complexity in the SSEVP-based experiment.

Drowsiness driving is one major factor of traffic accident. Monitoring the changes of brain signals provides an effective and direct way for drowsiness detection. One 3D convolutional neural network (3D CNN)-based forecasting system has been proposed to monitor electroencephalography (EEG) signals and predict fatigue level during driving. The limited weight sharing and channel-wise convolution were both applied to extract the significant phenomenon in various frequency bands of brain signals and the spatial information of EEG channel location, respectively. The proposed 3D CNN with limited weight sharing and channel-wise convolution has been demonstrated to predict reaction time (RT) of driving with low root mean square error (RMSE) through the brain dynamics. This proposed approach outperforms with the state-of-the-art algorithms, such as traditional CNN, Neural Network (NN), and support vector regression (SVR). Compared with traditional CNN and Artificial Neural Network, the RMSE of 3D CNN-based RT prediction has been improved 9.5% (RMSE from 0.6322 to 0.5720) and 8% (RMSE from 0.6217 to 0.5720), respectively. We envision that this study might open a new branch between deep learning application in neuro-cognitive analysis and real world application.

This correspondence paper proposes an improved version of the self-constructing neural fuzzy inference network (SONFIN), called soft-boosted SONFIN (SB-SONFIN). The design softly boosts the learning process of the SONFIN in order to decrease the error rate and enhance the learning speed. The SB-SONFIN boosts the learning power of the SONFIN by taking into account the numbers of fuzzy rules and initial weights which are two important parameters of the SONFIN, SB-SONFIN advances the learning process by: 1) initializing the weights with the width of the fuzzy sets rather than just with random values and 2) improving the parameter learning rates with the number of learned fuzzy rules. The effectiveness of the proposed soft boosting scheme is validated on several real world and benchmark datasets. The experimental results show that the SB-SONFIN possesses the capability to outperform other known methods on various datasets.

Aimed at the tremendous challenge of attribute reduction for big data mining and knowledge discovery, we propose a new attribute equilibrium dominance reduction accelerator (DCCAEDR) based on the distributed coevolutionary cloud model. First, the framework of N-populations distributed coevolutionary MapReduce model is designed to divide the entire population into N subpopulations, sharing the reward of different subpopulations' solutions under a MapReduce cloud mechanism. Because the adaptive balancing between exploration and exploitation can be achieved in a better way, the reduction performance is guaranteed to be the same as those using the whole independent data set. Second, a novel Nash equilibrium dominance strategy of elitists under the N bounded rationality regions is adopted to assist the subpopulations necessary to attain the stable status of Nash equilibrium dominance. This further enhances the accelerator's robustness against complex noise on big data. Third, the approximation parallelism mechanism based on MapReduce is constructed to implement rule reduction by accelerating the computation of attribute equivalence classes. Consequently, the entire attribute reduction set with the equilibrium dominance solution can be achieved. Extensive simulation results have been used to illustrate the effectiveness and robustness of the proposed DCCAEDR accelerator for attribute reduction on big data. Furthermore, the DCCAEDR is applied to solve attribute reduction for traditional Chinese medical records and to segment cortical surfaces of the neonatal brain 3-D-MRI records, and the DCCAEDR shows the superior competitive results, when compared with the representative algorithms.

In this paper, a novel approach is proposed to improve the classification performance of a polynomial neural network (PNN). In this approach, the partial descriptions (PDs) are generated at the first layer based on all possible combinations of two features of the training input patterns of a dataset. The set of PDs from the first layer, the set of all input features, and a bias constitute the chromosome of the real-coded genetic algorithm (RCGA). A system of equations is solved to determine the values of the real coefficients of each chromosome of the RCGA for the training dataset with the mean classification accuracy (CA) as the fitness value of each chromosome. To adjust these values for unknown testing patterns, the RCGA is iterated in the usual manner using simple selection, crossover, mutation, and elitist selection. The method is tested extensively with the University of California, Irvine benchmark datasets by utilizing tenfold cross validation of each dataset, and the performance is compared with various well-known state-of-the-art techniques. The results obtained from the proposed method in terms of CA are superior and outperform other known methods on various datasets.

This paper demonstrates a novel model for Mamdani type fuzzy inference system by using the knowledge learning ability of collaborative fuzzy clustering and rule learning capability of FCM. The collaboration process finds consistency between different datasets, these datasets can be generated at various places or same place with diverse environment containing common features space and bring together to find common features within them. For any kind of collaboration or integration of datasets, there is a need of keeping privacy and security at some level. By using collaboration process, it helps fuzzy inference system to define the accurate numbers of rules for structure learning and keeps the performance of system at satisfactory level while preserving the privacy and security of given datasets.

In this paper, a Takagi-Sugeno-Kang (TSK) type collaborative fuzzy rule based system is proposed with the help of knowledge learning ability of collaborative fuzzy clustering (CFC). The proposed method split a huge dataset into several small datasets and applying collaborative mechanism to interact each other and this process could be helpful to solve the big data issue. The proposed method applies the collective knowledge of CFC as input variables and the consequent part is a linear combination of the input variables. Through the intensive experimental tests on prediction problem, the performance of the proposed method is as higher as other methods. The proposed method only uses one half information of given dataset for training process and provide an accurate modeling platform while other methods use whole information of given dataset for training.

Preprocessing is generally used for data analysis in the real world datasets that are noisy, incomplete and inconsistent. In this paper, preprocessing is used to refine the inconsistency of the prototype and partition matrices before getting involved in the collaboration process. To date, almost all organizations are trying to establish some collaboration with others in order to enhance the performance of their services. Due to privacy and security issues they cannot share their information and data with each other. Collaborative clustering helps this kind of collaborative process while maintaining the privacy and security of data and can still yield a satisfactory result. Preprocessing helps the collaborative process by using an induced partition matrix generated based on cluster prototypes. The induced partition matrix is calculated from local data by using the cluster prototypes obtained from other data sites. Each member of the collaborating team collects the data and generates information locally by using the fuzzy c-means (FCM) and shares the cluster prototypes to other members. The other members preprocess the centroids before collaboration and use this information to share globally through collaborative fuzzy clustering (CFC) with other data. This process helps system to learn and gather information from other data sets. It is found that preprocessing helps system to provide reliable and satisfactory result, which can be easily visualized through our simulation results in this paper.

This study presents a fuzzy prediction system for the forecasting and estimation of driving fatigue, which utilizes a functional-link-based fuzzy neural network (FLFNN) to predict the drowsiness (DS) level in car driving task. The cognitive state in car driving task is one of key issue in cognitive neuroscience because fatigue driving usually causes enormous losses nowadays. The damage can be extremely decreased by the assistant of various artificial systems. Many Electroencephalography (EEG)-based interfaces have been widely developed recently due to its convenient measurement and real-time response. However, the improvement of recognition accuracy is still confined to some specific problems (e.g., individual difference). In order to solve this issue, the proposed methodology in this paper utilizes a nonlinear fuzzy neural network structure to increase the adaptability in the real-world environment. Therefore, this study is further to analysis the brain activities in car driving, which is constructed in a simulated three-dimensional virtual-reality (VR) environment. Finally, through the development of brain cognitive model in car driving task, this system can predict the cognitive state effectively before drivers' action and then provide correct feedback to users. This study also compared the result with the-state-of-art systems, including Linear Regression (LR), Multi-Layer Perceptron Neural Network (MLPNN) and Support Vector Regression (SVR). Results of this study demonstrate the effectiveness of the proposed FLFNN model.