Cognitive Computation最新文献

Although artificial intelligence has become part of everyone’s real life, a trust crisis against such systems is occurring, thus increasing the need to explain black-box predictions, especially in the military, medical, and financial domains. Modern eXplainable Artificial Intelligence (XAI) techniques focus on benchmark datasets, but the cognitive applicability of such solutions under big data settings is still unclear due to memory or computation constraints. In this paper, we extend a model-agnostic XAI methodology, named Cluster-Aided Space Transformation for Local Explanation (CASTLE), to be able to deal with high-volume datasets. CASTLE aims to explain the black-box behavior of predictive models by combining both local (i.e., based on the input sample) and global (i.e., based on the whole scope for action of the model) information. In particular, the local explanation provides a rule-based explanation for the prediction of a target instance as well as the directions to update the likelihood of the predicted class. Our extension leverages modern big data technologies (e.g., Apache Spark) to handle the high volume, variety, and velocity of huge datasets. We have evaluated the framework on five datasets, in terms of temporal efficiency, explanation quality, and model significance. Our results indicate that the proposed approach retains the high-quality explanations associated with CASTLE while efficiently handling large datasets. Importantly, it exhibits a sub-linear, rather than exponential, dependence on dataset size, making it a scalable solution for massive datasets or in any big data scenario.

This paper explores the convergence of Human-Centered AI (HCAI) and Cyber-Physical Social Systems (CPSS) in pursuing advanced Cognitive Situation Awareness (CSA). Integrating HCAI principles within CPSS fosters systems prioritizing human needs, values, and experiences, improving perception, understanding, and responsiveness to complex environments. By incorporating transparency, interpretability, and usability into Artificial Intelligence (AI) systems, the human-centered approach enhances user interaction and cooperation with intelligent systems, leading to more adaptive and efficient CPSS. The study employs a comprehensive approach to explore the intersection of HCAI and CPSS. Moreover, the paper presents case studies to illustrate real-world applications of HCAI and CPSS, such as self-driving cars and smart homes, transportation, healthcare, energy management, social media, and emergency response systems. Nevertheless, technical complexities, privacy concerns, and regulatory considerations must be addressed. The paper demonstrates the practical implications of integrating HCAI into CPSS through case studies in various domains. Furthermore, It highlights the positive impact of CSA systems such as self-driving cars, showcasing improvements in transportation. This paper contributes to advancing CSA and designing intelligent systems, promoting human–machine collaboration and societal well-being. By examining the intersection of HCAI and CPSS, this study advances research in CSA and designing intelligent systems prioritizing human needs, values, and experiences.

Fires represent a significant threat to the environment, infrastructure, and human safety, often spreading rapidly with wide-ranging consequences such as economic losses and life risks. Early detection and swift response to fire outbreaks are crucial to mitigating their impact. While satellite-based monitoring is effective, it may miss brief or indoor fires. This paper introduces a novel Perceived Risk Index (PRI) that, complementing satellite data, leverages social media data to provide insights into the severity of fire events. In the light of the results of statistical analysis, the PRI incorporates the number of fire-related tweets and the associated emotional expressions to gauge the perceived risk. The index’s evaluation involves the development of a comprehensive system that collects, classifies, annotates, and correlates social media posts with satellite data, presenting the findings in an interactive dashboard. Experimental results using diverse datasets of real-fire tweets demonstrate an average best correlation of 77% between PRI and the brightness values of fires detected by satellites. This correlation extends to the real intensity of the corresponding fires, showcasing the potential of social media platforms in furnishing information for emergency response and decision-making. The proposed PRI proves to be a valuable tool for ongoing monitoring efforts, having the potential to capture data on fires missed by satellites. This contributes to the development to more effective strategies for mitigating the environmental, infrastructural, and safety impacts of fire events.

Diagnosing cognitive impairment is an ongoing field of research especially in the elderly. Assessing the health status of the elderly can be a complex process that requires both subjective and objective measures. Subjective measures, such as self-reported responses to questions, can provide valuable information about a person’s experiences, feelings, and beliefs. However, from a scientific point of view, objective measures, based on quantifiable data that can be used to assess a person’s physical and cognitive functioning, are more appropriate and rigorous. The proposed system is based on the use of non-invasive instrumentation, which includes video images acquired with a frontal camera while the user performs different handwriting tasks on a Wacom tablet. We have acquired a new multimodal database of 191 elder subjects, which has been classified by human experts into healthy and cognitive impairment users by means of the standard pentagon copying test. The automatic classification was carried out using a video segmentation algorithm through the technique of shot boundary detection, in conjunction with a Transformer neural network. We obtain a multiclass classification accuracy of 77% and two-class accuracy of 83% based on frontal camera images, which basically detects head movements during handwriting tasks. Our automatic system can replicate human classification of handwritten pentagon copying test, opening a new method for cognitive impairment detection based on head movements. We also demonstrate the possibility to identifying the handwritten task performed by the user, based on frontal camera images and a Transformer neural network.

Cyber-physical-social-systems interconnect diverse technologies and communication infrastructure to the Internet for environmental sensing and computation. They offer many real-time autonomous services for smart cities, industry, transportation, medical systems, etc. The Internet of Medical Things (IoMT) has gained the potential for developing cyber-physical system (CPS) to facilitate healthcare applications and analyze the records of patients. Such a communication paradigm is integrated into many wireless standards for managing crucial data with cloud computing. However, the limitations of low-powered resources of such healthcare infrastructures increase the complexity level of sustainable growth. Wireless connectivity in next-generation networks is another research goal due to unbalanced load distribution. Furthermore, low-powered computing devices can be easily accessible by intruders and eliminate the confidentiality of any data transmission, so privacy is another research concern for healthcare systems. Therefore, using intelligent computing, this paper proposed a novel resilient predictive model for e-health sensing. The proposed model provides an efficient CPS-enabled automated routing system by exploring the optimization process with edge intelligence. This particular solution increases the level of cooperation between communication devices with intelligent data processing and higher predictive services. Moreover, by offering a trustworthy scheme, it seeks to enhance digital communication, data aggregation, and data breach prevention. The experimental findings highlight significant outcomes of the proposed model for packet reception, network overhead, data delay, and reliability as compared to alternative solutions.

Due to small imaging range of lens, blurring by jitter in the operation process and high similarity of urethral image features observed in different positions, doctors often face challenges in conducting a quick and comprehensive microscopic examination. In this paper, we combine image processing, simultaneous localization and mapping (SLAM) and intelligent navigation technologies to build an ORB-SLAM-based auxiliary microscopy guiding system. It can automatically process real-time microscopy videos, analyze the doctor’s detection path and provide direction for areas that have not been detected, assisting the doctor in completing urethral wall detection. In this system, a generative adversarial network-based deblurring algorithm is used to deblur the urethral images before SLAM processing. We creatively propose a vascular attention-based feature extraction algorithm tailored for urethral images. This algorithm combines F3Net and U-Net networks to detect the main body and branch points of blood vessels, respectively, which demonstrates the capability to assist the SLAM system in tracking the urethra more stably. Moreover, we design the direction guidance rules to aid doctors in urethral endoscopy. The system has been evaluated with a real urethral endoscope video dataset. Compared to other mainstream feature extraction algorithms, the method proposed in this paper is more accurate and comprehensive in identifying urethral vascular features, resulting in a 4.34% accuracy improvement, which confirms its effectiveness.

Diabetes mellitus (DM) can cause chronic foot issues and severe infections, including Diabetic Foot Ulcers (DFUs) that heal slowly due to insufficient blood flow. A recurrence of these ulcers can lead to 84% of lower limb amputations and even cause death. High-risk diabetes patients require expensive medications, regular check-ups, and proper personal hygiene to prevent DFUs, which affect 15–25% of diabetics. Accurate diagnosis, appropriate care, and prompt response can prevent amputations and fatalities through early and reliable DFU detection from image analysis. We propose a comprehensive deep learning-based system for detecting DFUs from patients’ feet images by reliably localizing ulcer points. Our method utilizes innovative model ensemble techniques—non-maximum suppression (NMS), Soft-NMS, and weighted bounding box fusion (WBF)—to combine predictions from state-of-the-art object detection models. The performances of diverse cutting-edge model architectures used in this study complement each other, leading to more generalized and improved results when combined in an ensemble. Our WBF-based approach combining YOLOv8m and FRCNN-ResNet101 achieves a mean average precision (mAP) score of 86.4% at the IoU threshold of 0.5 on the DFUC2020 dataset, significantly outperforming the former benchmark by 12.4%. We also perform external validation on the IEEE DataPort Diabetic Foot dataset which has demonstrated robust and reliable model performance on the qualitative analysis. In conclusion, our study effectively developed an innovative diabetic foot ulcer (DFU) detection system using an ensemble model of deep neural networks (DNNs). This AI-driven tool serves as an initial screening aid for medical professionals, augmenting the diagnostic process by enhancing sensitivity to potential DFU cases. While recognizing the presence of false positives, our research contributes to improving patient care through the integration of human medical expertise with AI-based solutions in DFU management.

Epileptic seizure prediction using multi-channel electroencephalogram (EEG) signals is very important in clinical therapy. A large number of channels lead to high computational complexity with low model performance. To improve the performance and reduce the overfitting that arises due to the use of unrelevant channels, the present paper proposed a channel selection method to study the brain region activation related to epileptic seizure. Our method is bio-inspired and cognitive since it integrates the novel binary many-objective particle swarm optimization with a ConvLSTM model. The proposed method has two advantages. First, it performed a new initialization strategy based on channel weighting with mutual information, thereby promoting the fast convergence of the optimization algorithm. Second, it captures spatio-temporal information from raw EEG segments thanks to the ConvLSTM model. The selected sub-channels are optimized as many-objective optimization problem that includes maximizing F1-score, sensitivity, specificity, and minimizing the ratio rate of selected channels. Our results have shown a performance of up to (97.94%) with only one EEG channel. Interestingly, when using all the EEG channels available, lower performance was achieved compared to the case when EEG channels were selected by our approach. This study revealed that it is possible to predict epileptic seizures using a few channels, which provides evidence for the future development of portable EEG seizure prediction devices.

Abstract

Hyperdimensional Computing (HDC), also known as Vector-Symbolic Architectures (VSA), is a promising framework for the development of cognitive architectures and artificial intelligence systems, as well as for technical applications and emerging neuromorphic and nanoscale hardware. HDC/VSA operate with hypervectors, i.e., neural-like distributed vector representations of large fixed dimension (usually > 1000). One of the key ingredients of HDC/VSA are the methods for encoding various data types (from numeric scalars and vectors to graphs) by hypervectors. In this paper, we propose an approach for the formation of hypervectors of sequences that provides both an equivariance with respect to the shift of sequences and preserves the similarity of sequences with identical elements at nearby positions. Our methods represent the sequence elements by compositional hypervectors and exploit permutations of hypervectors for representing the order of sequence elements. We experimentally explored the proposed representations using a diverse set of tasks with data in the form of symbolic strings. Although we did not use any features here (hypervector of a sequence was formed just from the hypervectors of its symbols at their positions), the proposed approach demonstrated the performance on a par with the methods that exploit various features, such as subsequences. The proposed techniques were designed for the HDC/VSA model known as Sparse Binary Distributed Representations. However, they can be adapted to hypervectors in formats of other HDC/VSA models, as well as for representing sequences of types other than symbolic strings. Directions for further research are discussed.

In identifying objects, people usually associate memory templates to guide visual attention and determine the category of an object. The initial character images that children learn are usually normal patterns. However, the variation in corresponding handwritten patterns is quite large. To learn these deformed images with large variance, current deep models must involve millions of parameters for such kind of classification tasks that seem much easier and simpler to children who learn to recognize new characters associated with their initially taught normal patterns. From the perspective of humans’ perception, when people see a new object, they first think of a template image in their memory, which is similar to the object. This mapping process makes it easier for humans to learn new objects. Inspired by this cognitive association mechanism, this study developed a cognition-inspired handwritten character recognition model using a proposed normal template mapping neural network. This model uses an encoder-decoder architecture to build a normal template mapping neural network that transforms handwritten character images of one class to normalized characters similar to a given printed template character image representing that class. Then, a simple shallow classifier recognizes these normalized images, which are easier to classify. The experimental results show that the proposed model completes handwritten character recognition with comparable or higher precision at a much lower parameter count than current representative deep models. The proposed model removes the individual styles of handwritten character images and maps them to patterns similar to normal template images. This greatly reduces the classification difficulty and enables the classifier to classify only known standard character images.