Cognitive Computation

Food spoilage detection is critical in ensuring food safety and reducing waste. In this work, we offer a new neural network model, rotOrNot, intended for image analysisbased rotten food detection. Our method focuses on accurately identifying spoiled food, precisely spotting rot. The proposed method meets a vital demand for reducing food waste and guaranteeing food safety. Through several performance criteria, the proposed model was thoroughly tested and showed better performance than current approaches. rotOrNot specifically obtained training accuracy of 99.4%, validation accuracy of 96.95%, an outstanding ROC-AUC score of 99.97296%, and a test accuracy of 97.52%. Moreover, rotOrNot shows the fastest training time and exceeds the previous models in accuracy, so it is a quite effective tool for practical usage. This work offers a useful instrument for the automatic food rotting detection, therefore improving food safety procedures and maybe reducing food waste.

In modern agriculture, ensuring plant health is essential for high crop yields and quality. Plant diseases pose risks to economies, communities, and the environment, making early and accurate diagnosis crucial. The internet of things (IoT) has revolutionized farming by enabling real-time crop monitoring and using drones and cameras for early disease detection. This technology helps farmers address challenges with precision and sustainability. This research proposes an ensemble learning model incorporating multi-class capsule networks (MCCN) and other pre-trained model with majority voting system is implemented to predict plant diseases and pests early. The research aims to develop a robust MCCN-based ensemble prediction model for timely disease identification. To evaluate the performance of the ensemble model, various key metrics, including accuracy, and loss value, are assessed. Furthermore, a comparative analysis is conducted, benchmarking the MCCN model against other well-known pre-trained models such as residual network-101 (ResNet101), visual geometry group-19 (VGG19), and GoogleNet. This research signifies a substantial stride towards the realization of IoT-driven precision agriculture, where advanced technology and machine learning contribute to the early detection and mitigation of plant diseases, ultimately enhancing crop yield and environmental sustainability.

We present a novel language-independent technique for determining polarity, positive or negative, of opinions expressed by different individuals. The technique is based on byte-level n-gram frequency statistics method for document representation, and a variant of k nearest neighbors (kNN) (for k = 1) machine learning algorithm for categorization process. The main advantages of the technique are its simplicity and full language and topic independence. For driving experiments we used corpora of movie reviews: Cornell polarity dataset in English and MuchoCine in Spanish. Experimental results (85.6% accuracy for English and 82.49% for Spanish corpora) confirm that the presented technique is comparable with the best ranked previously published techniques, when applied to movie reviews datasets. Still, it use no additional linguistic information nor external resources.

Diabetic Foot Ulcer (DFU) is a multifarious disease of diabetes
characterized by wound infection in the feet and thus creates a public
health issue. Initially, it is a superficial wound in the feet, but it goes deeper and moves toward the upper portion. Currently, 18.6 million
people are affected by DFU, and the number is increasing due to different
contributing factors. Due to its high prevalence, immediate prevention
and treatment are highly recommended. Various classification systems,
each with pros and cons, are proposed which may help the healthcare
providers to recognize, prevent, and manage the DFU effectively.
Different treatment options are available but have certain limitations and
challenges that need to be addressed by advanced and automated
technologies. In the modern technological era, the role of artificial
intelligence and its allies got exceptional attention among researchers.
The role of AI in healthcare settings, particularly in the prevention and
management of disease, is its modern and latest applications. In the realm
of disease treatment potential, the role of AI tools in DFU detection,
prevention, and management has flourished well. These tools can help
clinicians and doctors in many ways in the spectrum of DFU. For
instance, early detection, image recognition, and big data analysis are a
few glimpses. Besides, the management, AI and machine learning tools
can also help in decision-making. Role in DFU diagnosis, identification
of potential DFU sites, and development of personalized equipment for
the prevention and treatment of DFU are their advanced applications.
Currently, the AI tools in the different spectrums of DFU are used
successfully, although limitations of privacy, low experimental results,
and user compatibility are still underlined. This chapter sheds light on
DFU, its classification system, and fully describes the role of AI tools in
the detection, prevention, management, and treatment of DFU.

Brain Tumors (BT) are one of the most hazardous diseases in the world, primarily because of the speed at which they spread and high mortality rates. Hence, diagnosing BT at an early stage with higher accuracy is necessary. Recent breakthroughs in deep transfer learning have achieved great performances in identifying common types of BT, such as glioma, meningioma, and pituitary tumors. In this paper, we will propose a novel hybrid deep transfer learning architecture by combining the strengths of two stateof-the-art pre-trained deep convolutional neural networks: MobileNetV2 and InceptionV3. We implement a parallel feature extraction process, reduce dimensionality in the extracted features using dropout techniques, and pass the refined data through fully connected layers for accurate classification. In this context, our two-step methodology of sub-image classification and final image categorisation with the help of majority voting and maximum probability principles shows much-improved results. Validated on a comprehensive BT dataset, our approach achieves 98.68% accuracy, 97% precision, 97.25% recall, and 97% F-measure with 0.9975 AUC and average precision of 99.50%. The outcomes surpass the performance of existing models on precision, sensitivity, and generalization, thus revealing the value that this technique may hold in whole-slide MRI image classification and clinical practice.

Cyberbullying is a pressing issue in the digital age, especially in non-English-speaking communities. This study focuses on using deep learning techniques to detect cyberbullying in the Bangla language. The research employs a Bangla dataset that includes both cyberbullying and noncyberbullying text, which undergoes preprocessing, tokenization, and sequence transformation for input into a LSTM recurrent neural network. The LSTM model incorporates word embedding and dropout techniques to enhance its performance. Evaluation metrics like accuracy and AUC-ROC score are employed to assess the model's effectiveness. The study investigates many methods, such as vernacular-specific preparation and data augmenting, that are specifically designed for the identification of cyberbullying in the Bangla language. The findings demonstrate deep learning's capacity to identify cyberbullying in Bangla, which makes a significant contribution to resolving this problem in non-English settings. The research reveals impressive accuracy rates, with LSTM achieving 99.99% accuracy, and a different approach, Convolutional Neural Network (CNN), yielding 99.022%. Ensemble models, including Gradient Boosting, XGBoost, and Random Forest, are also employed and optimized using hyperparameter tuning. Notably, LSTM stands out as the most reliable classifier for cyberbullying detection, as confirmed by the experimental findings and recent research.

Researchers have studied Stewart-Gough platforms, also known as Gough-Stewart platforms or hexapod platforms extensively for their inherent fine control characteristics. Their studies led to the potential deployment opportunities of Stewart-Gough Platforms in many critical applications such as the medical field, engineering machines, space research, electronic chip manufacturing, automobile manufacturing, etc. Some of these applications need micro and nano-level movement control in 3D space for the motions to be precise, complicated, and repeatable; a Stewart-Gough platform fulfills these challenges smartly. For this, the platform must be more accurate than the specified application accuracy level and thus proper calibration for a parallel robot is crucial. Forward kinematicsbased calibration for these hexapod machines becomes unnecessarily complex and inverse kinematics complete this task with much ease. To experiment with different calibration techniques, various calibration approaches were implemented by using external instruments, constraining one or more motions of the system, and using extra sensors for auto or self-calibration. This survey paid attention to those key methodologies, their outcome, and important details related to inverse kinematic-based parallel robot calibrations. It was observed during this study that the researchers focused on improving the accuracy of the platform position and orientation considering the errors contributed by one source or multiple sources. The error sources considered are mainly kinematic and structural, in some cases, environmental factors also are reviewed, however, those calibrations are done under no-load conditions. This study aims to review the present state of the art in this field and highlight the processes and errors considered for the calibration of Stewart-Gough platforms.

Background/introduction: The nonferrous metallurgy industry is a major energy consumer in China, and accurate energy consumption forecasting for the nonferrous metallurgy industry can help government policymakers with energy planning. Methods: For this purpose, a hybrid support vector regression (HSVR) with an adaptive state transition algorithm (ASTA) named ASTA-HSVR is proposed to forecast energy consumption in the nonferrous metallurgy industry. The proposed support vector regression (SVR) model consists of a linear weighting of ϵ-SVR and ν-SVR. The ASTA was developed to optimize the parameters of the HSVR. Results: Two cases of energy consumption from the nonferrous metallurgy industry in China are used to demonstrate the performance of the proposed method. The results indicate that the ASTA-HSVR method is superior to other methods.

Alzheimer's Disease (AD) is a progressive neurodegenerative disorder that impairs memory, cognition, and daily functioning, often accompanied by significant behavioral and personality changes in older adults. While there is currently no cure, treatment interventions are most effective during the early and middle stages, emphasizing the critical need for early diagnosis-especially as global aging trends contribute to a rising prevalence of AD. This study aims to enhance early detection of Alzheimer's Disease by leveraging machine learning and deep learning techniques on MRI scans from the Open Access Series of Imaging Studies (OASIS) dataset. We employed a diverse set of models, including Random Forest and Logistic Regression from classical machine learning, Extra Trees from ensemble learning, and Convolutional Neural Networks (CNNs) from deep learning. Performance was assessed using accuracy, precision, recall, and area under the curve (AUC). Results indicate that CNNs achieved the highest accuracy and AUC scores, while Extra Trees excelled in precision and recall-highlighting the potential of both deep learning and ensemble-based methods in supporting early AD diagnosis through neuroimaging.

chronic kidney disease (CKD) is often diagnosed at later stages, leading to severe health impacts. This study presents a machine learning-based approach for early CKD prediction using patient clinical data. To improve model transparency, Explainable AI (XAI) techniques like LIME are employed, offering insights into feature contributions. The proposed system achieves high accuracy and supports clinical decision-making by identifying key indicators influencing CKD onset.

Sentiment analysis is an emerging research area and is very important because human beings are largely dependent on the web in these days. A little research on sentiment analysis has been done for Myanmar language in Myanmar. So, this paper aims to review past, present and future of Myanmar sentiment analysis and presents the study of the Myanmar language research on the opinion lexicon creation and sentiment classification. The lexicons are created for four domains and the sentiment classification is done for three domains by using the equations of polarity scores, Support Vector Machine (SVM), Long Short-Term Memory (LSTM) and Naïve Bayes. The objective of this paper is to provide a review of Myanmar language sentiment analysis research papers from well-known journals and conferences between 2017 and 2019.

Objective: Gambling fallacies are believed to be etiologically related to the development of problem gambling. However, this evidence is tenuous due to the lack of consensus on which things constitute gambling fallacies and the adequacy of instruments that ostensibly measure them. The purpose of this paper is to comprehensively identify the main gambling fallacies and examine the reliability and validity of the instruments designed to measure them. Methods: All known gambling fallacies and instruments measuring them were identified via a keyword search of social science, medical, and gambling-specific databases. The reliability and validity of each assessment instrument was then examined. Results: Six primary gambling fallacies were consistently reported in the literature. Eighteen instruments were found to measure one or more of these fallacies, with 9 assessing specific fallacies and 9 intended to be comprehensive instruments. Most instruments were found to have good internal consistency as well as adequate convergent and external validity. Relatively few demonstrated test-retest reliability and/or discriminant validity. However, the main area of concern was content validity. While instruments focusing on a particular fallacy tended to have adequate content validity, this was not true of the comprehensive instruments. In addition to insufficient coverage of the fallacies, most comprehensive instruments included questions pertaining to motivations for gambling, attitudes about gambling, and/or problem gambling symptomatology (e.g. chasing losses), which likely inflates their statistical association with problem gambling. Many of these comprehensive instruments also wrongly assume that no skill is involved in any form of gambling. The inadequate content validity of most comprehensive gambling fallacy instruments draws into question the strong etiological relationship gambling fallacies are presumed to have with problem gambling. This concern is compounded by the fact that all research reporting this association has been cross-sectional and correlational in nature. Re-examination of this relationship using improved instrumentation in a longitudinal context is required.

As recommended by the international standards, ISO 21542, ease of wayfinding must be ensured by installing signage at all key decision points on walkways such as forks because signage greatly influences the way in which people unfamiliar with an environment navigate through it. Therefore, we aimed to develop a new system for evaluating the ease of wayfinding, which could detect spots that cause disorientation, i.e., "disorientation spots", based on simulated three-dimensional (3D) interactions between wayfinding behaviors and signage location, visibility, legibility, noticeability, and continuity. First, an environment model reflecting detailed 3D geometry and textures of the environment, i.e., "as-is environment model", is generated automatically using 3D laser-scanning and structure-from-motion (SfM). Then, a set of signage entities is created by the user. Thereafter, a 3D wayfinding simulation is performed in the as-is environment model using a digital human model (DHM), and disorientation spots are detected. The proposed system was tested in a virtual maze and a real two-story indoor environment. It was further validated through a comparison of the disorientation spots detected by the simulation with those of six young subjects. The comparison results revealed that the proposed system could detect disorientation spots, where the subjects lost their way, in the test environment.

Alzheimer's disease is the most prevalent etiology of dementia, and its early diagnosis has been deemed instrumental for early intervention and better patient management. Over the last few years, deep learning techniques have been used with neuroimaging data, and notable gains in diagnostic performance have ensued. Such models have, however, typically been faulted for being "black boxes" that, despite their performance, hinder their acceptance in clinical practice due to a lack of transparency and interpretability. Current explainability methods have typically been used as post-hoc procedures, but they often yield inconsistent or anatomically irrelevant attribution maps that clinicians find hard to trust. In this review, progress in explainable AI for the detection of Alzheimer's is discussed, with a focus on DL methods combined with multimodal neuroimaging. Emphasis is given to XRAI, a region-based attribution method that has been demonstrated to produce more coherent and clinically interpretable explanations compared to conventional pixel-level techniques. The application of XRAI to 2D and 3D neuroimaging is considered, together with the potential of XRAI to identify anatomically relevant brain areas implicated in disease pathology. Challenges in terms of clinical uptake, integration into workflow, and standardized evaluation of explainability techniques are also discussed. The review identifies how the pairing of high-performing AI models with strong explainability methods has the potential to enable the creation of practical and reliable diagnostic tools for real-world clinical application.

Financial economic research has extensively documented the fact that the impact of the arrival of negative news on stock prices is more intense than that of the arrival of positive news. The authors of the present study followed an innovative approach based on the utilization of two artificial intelligence algorithms to test that asymmetric response effect. Methods: The first algorithm was used to web-scrape the social network Twitter to download the top tweets of the 24 largest market-capitalized publicly traded companies in the world during the last decade. A second algorithm was then used to analyze the contents of the tweets, converting that information into social sentiment indexes and building a time series for each considered company. After comparing the social sentiment indexes’ movements with the daily closing stock price of individual companies using transfer entropy, our estimations confirmed that the intensity of the impact of negative and positive news on the daily stoc...

🧠 Abstract
This paper presents a complete implementation of a facial recognition system using artificial intelligence (AI) technologies. The goal of this project is to design, develop, and test a practical face recognition model capable of identifying individuals in real-time from image or video input. The system integrates several tools and libraries such as Python, OpenCV, face_recognition, and Flask for web deployment. The results of the project demonstrate how facial recognition can be implemented efficiently for academic, security, and identity verification applications.

Systems developed in wearable devices with sensors on board are widely used to collect data of of humans and animals activities with the prospective of an on-board automatic classification of data. An interesting application of these systems is to support animals' behaviour monitoring gathered by sensors' data analysis. This is a challenging area and in particular with fixed memories capabilities because the devices should be able to operate autonomously for long periods before being retrieved by human operators, and being able to classify activities on board can significantly improve their autonomy. In this paper we focus on the identification of prey handling activity in seals (when the animal start attaching and biting the prey), which is one of the main movement that identify a successful foraging activity. Data taken into consideration are streams of 3D accelerometers and depth sensors values collected by devices attached directly on seals. To analyse these data, we propose an automatic model based on Machine Learning (ML) algorithms. In particular we compare the performance (in terms of accuracy and F1score) of three ML algorithms: Input Delay Neural Networks, Support Vector Machines, and Echo State Networks. We attend to the final aim of developing an automatic classifier on-board. For this purpose, in this paper the comparison is performed concerning the performance obtained by each ML approach developed and its memory footprint. In the end, we highlight the advantage of using an ML algorithm, in terms of feasibility in wild animals' monitoring.

Sound synthesis plays a central role in the compositional process in the academic field of electroacoustic domain. This research aims to investigate and integrate the use of generative artificial intelligence models as tools for sound synthesis. Three models, SynthIo, Mu-sicLM and MusicGen, were used in this study. Ten prompts were designed and tested with the aim of generating sound textures. To assess the consistency of the generated samples, three expert electroacoustic music composers evaluated the different samples related to specific requirements. CCS Concepts • Information systems → User Modelling Applications, Wearable Devices.

Multi-agent reinforcement learning introduces unique challenges in cooperative environments, particularly when agents must coordinate complex, interdependent actions in environments with sparse rewards. The Overcooked-AI benchmark exemplifies such challenges by simulating constrained kitchen environments where two agents must collaborate to efficiently prepare and deliver soups. This paper investigates the use of a centralized training and decentralized execution (CTDE) framework to enhance coordination between agents.

A new hypothesis is proposed, that early life possessed a mechanism for the simultaneous synthesis of a polypeptide and its coding mRNA. Early tRNAs and mRNAs are considered to have been pairs of simple complementary nucleotide triplets, able to catalyze peptide bond formation and joining of the codon triplets. An in silico build stereochemical model for such a process is presented. A model for the evolution of modern tRNAs and ribosomal translation from such a primitive apparatus is also presented. Modern tRNAs are viewed as consisting of two halves, one of which evolved from ancient elongating mRNA and the other from a primitive also elongating tRNA. The probable origin of the ACC triplet at the amino-acid-accepting arm of the tRNAs is discussed. A theoretical scheme for the generation of mutations in response to the presence of chemical or physical factors and to their effect on newly synthesized polypeptides, based on the polypeptide-mRNA co-synthesis hypothesis, is presented. Messenger RNA nicking during each step of peptide bond formation and trans-translation of a different mRNA or codon triplet, in response to ribosome stalling, are the essential elements of the proposed theory for directed mutability.

This paper presents a Bayesian framework for under-determined audio source separation in multichannel reverberant mixtures. We model the source signals as Student's t latent random variables in a time-frequency domain. The specific structure of musical signals in this domain is exploited by means of a non-negative matrix factorization model. Conversely, we design the mixing model in the time domain. In addition to leading to an exact representation of the convolutive mixing process, this approach allows us to develop simple probabilistic priors for the mixing filters. Indeed, as those filters correspond to room responses they exhibit a simple characteristic structure in the time domain that can be used to guide their estimation. We also rely on the Student's t distribution for modeling the impulse response of the mixing filters. From this model, we develop a variational inference algorithm in order to perform source separation. The experimental evaluation demonstrates the potential of this approach for separating multichannel reverberant mixtures.

Recent developments of sensors that allow tracking of human movements and gestures enable rapid progress of applications in domains like medical rehabilitation or robotic control. Especially the inertial measurement unit (IMU) is an excellent device for real-time scenarios as it rapidly delivers data input. Therefore, a computational model must be able to learn gesture sequences in a fast yet robust way. We recently introduced an echo state network (ESN) framework for continuous gesture recognition (Tietz et al., 2019) including novel approaches for gesture spotting, i.e., the automatic detection of the start and end phase of a gesture. Although our results showed good classification performance, we identified significant factors which also negatively impact the performance like subgestures and gesture variability. To address these issues, we include experiments with Long Short-Term Memory (LSTM) networks, which is a state-of-the-art model for sequence processing, to compare the obt...

Emotion-cause pair extraction (ECPE) task aims to extract all the pairs of emotions and their causes from an unannotated emotion text. The previous works usually extract the emotion-cause pairs from two perspectives of emotion and cause. However, emotion extraction is more crucial to the ECPE task than cause extraction. Motivated by this analysis, we propose an end-to-end emotion-cause extraction approach oriented toward emotion prediction (EPO-ECPE), aiming to fully exploit the potential of emotion prediction to enhance emotion-cause pair extraction. Considering the strong dependence between emotion prediction and emotion-cause pair extraction, we propose a synchronization mechanism to share their improvement in the training process. That is, the improvement of emotion prediction can facilitate the emotion-cause pair extraction, and then the results of emotion-cause pair extraction can also be used to improve the accuracy of emotion prediction simultaneously. For the emotion-cause pair extraction, we divide it into genuine pair supervision and fake pair supervision, where the genuine pair supervision learns from the pairs with more possibility to be emotion-cause pairs. In contrast, fake pair supervision learns from other pairs. In this way, the emotion-cause pairs can be extracted directly from the genuine pair, thereby reducing the difficulty of extraction. Experimental results show that our approach outperforms the 13 compared systems and achieves new state-of-the-art performance.

Emotion-cause pair extraction (ECPE) task aims to extract all the pairs of emotions and their causes from an unannotated emotion text. The previous works usually extract the emotion-cause pairs from two perspectives of emotion and cause. However, emotion extraction is more crucial to the ECPE task than cause extraction. Motivated by this analysis, we propose an end-to-end emotion-cause extraction approach oriented toward emotion prediction (EPO-ECPE), aiming to fully exploit the potential of emotion prediction to enhance emotion-cause pair extraction. Considering the strong dependence between emotion prediction and emotion-cause pair extraction, we propose a synchronization mechanism to share their improvement in the training process. That is, the improvement of emotion prediction can facilitate the emotion-cause pair extraction, and then the results of emotion-cause pair extraction can also be used to improve the accuracy of emotion prediction simultaneously. For the emotion-cause pair extraction, we divide it into genuine pair supervision and fake pair supervision, where the genuine pair supervision learns from the pairs with more possibility to be emotion-cause pairs. In contrast, fake pair supervision learns from other pairs. In this way, the emotion-cause pairs can be extracted directly from the genuine pair, thereby reducing the difficulty of extraction. Experimental results show that our approach outperforms the 13 compared systems and achieves new state-of-the-art performance.

We carried out a series of experiments on text classification using multi-word features. A hand-crafted method was proposed to extract the multi-words from text data set and two different strategies were developed to normalize the multi-words into two different versions of multi-word features. After the texts were represented respectively using these two different multi-word features, text classification was conducted in contrast to examine the effectiveness of these two strategies. Also the linear and nonlinear polynomial kernel of support vector machine (SVM) was compared on the performance of text classification task.

Automatic Text Summarization strategies have been successfully employed to digest text collections and extract its essential content. Usually, summaries are generated using textual corpora that belongs to the same domain area where the summary will be used. Nonetheless, there are special cases where it is not found enough textual sources, and one possible alternative is to generate a summary from a different domain. One manner to summarize texts consists of using a graph model. This model allows giving more importance to words corresponding to the main concepts from the target domain found in the summarized text. This gives the reader an overview of the main text concepts as well as their relationships. However, this kind of summarization presents a significant number of repeated terms when compared to human-generated summaries. In this paper, we present an approach to produce graph-model extractive summaries of texts, meeting the target domain exigences and treating the terms repet...

People above the age of 65 are disproportionately affected by "Alzheimer's disease (AD)". AD is one of the most prevalent forms of dementia that causes a gradual decline in their cognition and mental abilities. Immediate medication is essential to halt the course of dementia, but this relies on an accurate AD diagnosis and its prior stage, "Mild Cognitive Impairment (MCI)". Through this effect, an accurate assessment from neuroimaging is needed. An effective diagnostics approach supported by the processing of "Magnetic Resonance Imaging (MRI)" results might provide a better comprehensive and trustworthy solution, and may even improve diagnostic accuracy. In the past, researchers examining MRI scans for AD problems have relied on massive multivariate analysis, which assumes that individual brain areas function autonomously. However, this view is inappropriate for comprehending how far the brain functions. In this research, an "Advanced Adaptive Neuro-Fuzzy Inference System (AANFIS)" classification is proposed for automatic multi-level AD categorization through MRI brain images. The "Neural Network (NN)" is used to enhance the roles of fuzzy membership as decision-making structures for the given MRI brain image. The "Fuzzy Logic (FL)" seems to have the capacity to explicitly transmit expertise results. If this is the case, it tries to enforce laws. However, it takes too long to establish and administer the Member States' position, which perceives certain linguistic elements quantitatively. Whereas automated learning is possible in NN processes. It lowers the expense and time of instruction. Also, it increases the performance of the classifier. So the combination of Advanced NN with FL offers stronger outcomes for AD classification. Classifying subjects of sMRI brain scans into "Cognitive Normal (CN)", "Mild Cognitive Impairment (MCI)", and "Pure Alzheimer's Disease (AD)" categories have been the focus of this present research, which employs an AANFIS classification. The proposed AANFIS method outperforms the state-of-the-art KNN as well as EFKNN methods in aspects of classifying and identification performance.

A less studied component of gaze allocation in dynamic real-world scenes is the time lag of eye movements in responding to dynamic attention-capturing events. Despite the vast amount of research on anticipatory gaze behaviour in natural situations, such as action execution and observation, little is known about the predictive nature of eye movements when viewing different types of natural or realistic scene sequences. In the present study, we quantify the degree of anticipation during the free viewing of dynamic natural scenes. The cross-correlation analysis of image-based saliency maps with an empirical saliency measure derived from eye movement data reveals the existence of predictive mechanisms responsible for a nearzero average lag between dynamic changes of the environment and the responding eye movements. We also show that the degree of anticipation is reduced when moving away from natural scenes by introducing camera motion, jump cuts, and film-editing.

Predictive business process monitoring (PBPM) provides a set of techniques to perform different predic­ tion tasks in running business processes, such as the next activity, the process outcome, or the remaining time. Nowadays, deep­learning­based techniques provide more accurate predictive models. However, the explainability of these models has long been neglected. The predictive quality is essential for PBPM­ based decision support systems, but also its explainability for human stakeholders needs to be considered. Explainable artificial intelligence (XAI) describes different approaches to make machine­learning­based techniques explainable. To examine the current state of explainable PBPM techniques, we perform a structured and descriptive literature review. We identify explainable PBPM techniques of the domain and classify them along with different XAI­related concepts: prediction purpose, intrinsically interpretable or post­hoc, evaluation objective, and evaluation method. Based o...

In this paper, the aim is to build a hybrid Word Sense Disambiguation(WSD) technique, which is acutely focused on text associated with a certain form of visual. Natural language processing helps establish a context among the data elements that are aggregated to establish a certain meaning. Analyzing transcripts of visuals being uploaded in real-time saves resources and time required to sort content based on genres or emotions. The training data lays a foundation to rate the polarities of elements, on top of which the dictionary expands as an when new content is supplied to the apparatus. Third-party intelligence is combined with the dictionary to experience growth even when the consumer usage is idle. All these entities are mutually intertwined to ensure maximum utility and output.

For AI solutions to be more reliable and effective, Human-Computer Interaction (HCI) and Explainable Artificial Intelligence (XAI) must come together in the healthcare industry. Support Vector Classifier (SVC), Sequential models, Random Forest Classifier, GaussianNB, and K-Nearest Neighbours (KNN) are just a few of the AI models that are thoroughly reviewed in this work, with an emphasis on their interpretability and performance. The models with the best accuracy, SVC (0.8771), were followed by KNN (0.7881), GaussianNB (0.8474), Random Forest (0.8517), and Sequential models (0.8559). The study highlights the significance of openness and intuitive AI systems in promoting their adoption by medical professionals. Model-agnostic explanation methods like as SHAP and LIME are recognised as key instruments for enhancing interpretability. The review emphasises how important it is to strike a balance between explainability and AI model performance in order to guarantee trustworthy and understandable AI applications in healthcare contexts.

Background Information: The management of chronic diseases, fall prevention, and proactive healthcare are essential for enhancing care for the ageing population. Artificial intelligence (AI) and machine learning (ML) provide sophisticated instruments for predictive modelling, facilitating prompt interventions and tailored treatment approaches in geriatric care. Objectives: The objective of this work is to create predictive models utilising artificial intelligence and machine learning for the management of chronic diseases, fall detection, and preventative healthcare applications, hence improving care quality and patient outcomes in elderly populations. Methods: Logistic Regression, Random Forest, and Convolutional Neural Network (CNN) models were trained using clinical and sensor data, both individually and in ensemble combinations, to forecast health risks. Results: The ensemble model attained 92% accuracy, 90% precision, 89% recall, 90% F1score, and 91% AUC-ROC, indicating superior predictive performance across all criteria. Conclusion: In conclusion, ensemble-based AI models improve risk prediction for the aged, facilitating proactive treatments and enhancing healthcare outcomes for senior patients.

Brain  tumorspose  a  significant  challenge  in  medical  diagnostics,  requiring  early  and  accurate  detection  for effective  treatment.  Traditional  diagnostic  methods rely heavily  on  manual  assessment  of  MRI  scans,  which  can  be  time-consuming  and  prone  to  human  error.  Recent  advancements  in  deep  learning  and  cloud  computing  have  enabled  the development of automated, scalable, and high-accuracy medical imaging solutions.This study presents an AI-Driven Cloud-Based  Deep  Learning  Framework  for predictive  healthcare  analytics, integrating  Convolutional  Neural  Networks  (CNNs) with cloud computing to enhance scalability and computational efficiency. The Brain Tumor MRI Dataset from Kaggle is used,  and  extensive  data  preprocessing  techniques  such  as  noise  reduction,  normalization, and  contrast  enhancement  are applied  to  improve  image  quality.  The  proposed  model  achieves  97.2%  accuracy,  outperforming  traditional  methods  like SVM  (88.5%)  and  Random  Forest  (91.3%),  with  an  AUC-ROC  score  of  98.3%,  demonstrating  superior  classification capability.  Cloud-based  deployment  significantly  reduces  training  time  to  45  minutes  compared  to  over  3  hours  on  local systems, ensuring rapid and resource-efficient processing of large MRI datasets. The findings highlight the potential of AI-driven,  cloud-integrated  deep  learning  for  real-time,  scalable,  and  high-accuracy  medical  diagnostics,  offering  a transformative  approach  to  brain  tumor  detection  in  predictive  healthcare  analytics.The  proposed  AI-driven  cloud-based deep  learning  framework  achieves  high  accuracy  in  brain  tumor  classification  while  significantly  reducing  training  time through  scalable  cloud  computing.  This  approach  enhances  diagnostic  efficiency,  making  advanced  medical  imaging solutions more accessible and effective.

Online handwritten analysis presents many applications in e-security, signature biometrics being the most popular but not the only one. Handwriting analysis also has an important set of applications in e-health. Both kinds of applications (e-security and e-health) have some unsolved questions and relations among them that should be addressed in the next years. We summarize the state of the art and applications based on handwriting signals. Later on, we focus on the main achievements and challenges that should be addressed by the scientific community, providing a guide for future research. Among all the points discussed in this article, we remark the importance of considering security, health, and metadata from a joint perspective. This is especially critical due to the risks inherent when using these behavioral signals.

Human creativity is personally, socially and culturally situated: creative individuals work within environments rich in personal experiences, social relationships and cultural knowledge. Computational models of creative processes typically neglect some or all of these aspects of human creativity. How can we hope to capture this richness in computational models of creativity? This paper introduces recent work at the Design Lab where we are attempting to develop a model of artificial creative systems that can combine important aspects at personal, social and cultural levels.

The fundamental, powerful process of computation in the brain has been widely misunderstood. The paper associates the general failure to build intelligent thinking machines with current reductionist principles of temporal coding and advocates for a change in paradigm regarding the brain analogy. Since fragments of information are stored in proteins which can shift between several structures to perform their function, the biological substrate is actively involved in physical computation. The intrinsic nonlinear dynamics of action potentials and synaptic activities maintain physical interactions within and between neurons in the brain. During these events the required information is exchanged between molecular structures (proteins) which store fragments of information and the generated electric flux which carries and integrates information in the brain. The entire process of physical interaction explains how the brain actively creates or experiences 'meaning'. This process of interaction during an action potential generation can be simply seen as the moment when the neuron solves a many-body problem. A neuroelectrodynamic theory shows that the neuron solves equations rather than exclusively computes functions. With the main focus on temporal patterns, the spike timing dogma (STD) has neglected important forms of computation which do occur inside neurons. In addition, artificial neural models have missed the most important part since the real super-computing power of the brain has its origins in computations that occur within neurons.

Natural systems can provide excellent solutions to build artificial intelligent systems. The brain represents the best model of computation that leads to general intelligent action. However, current mainstream models reflect a weak understanding of computations performed in the brain that is translated in a failure of building powerful thinking machines. Specifically, temporal reductionist neural models elude the complexity of information processing since spike timing models reinforce the idea of neurons that compress temporal information and that computation can be reduced to a communication of information between neurons. The active brain dynamics and neuronal data analyses reveal multiple computational levels where information is intracellularly processed in neurons. New experimental findings and theoretical approach of neuroelectrodynamics challenge current models as they now stand and advocate for a change in paradigm for bioinspired computing machines.

Although existing denoising techniques such as Convolutional Neural Networks (CNNs) [Zhang et al., 2017] and Multilayer Perceptrons (MLPs) [Burger et al., 2012] have made remarkable progress, they still face challenges in handling noise effectively. Recent advances in image denoising through deep learning-based models such as FFDNet and DnCNN [Zhang et al., 2018] have shown significant promise in noise reduction. However, these methods often fall short in effectively handling encrypted images due to limitations in preserving fine image details under diverse noise scenarios [Dong et al., 2019; Li et al., 2023]. The core focus of this paper is on addressing the unique complexities of denoising encrypted images. Unlike conventional methods that struggle with preserving image details, our model leverages deeper network architectures and advanced deep learning techniques, particularly the Deep Convolutional Residual Network (Deep ConvResNet), to achieve superior denoising performance. ResNet-based methods [Ren et al., 2019; Zhang et al., 2021] not only efficiently denoise images but also defend against noise, occlusion, and blur attacks.

In today's fast-paced lifestyle, pursuing holistic well-being has increased interest in monitoring and managing stress levels. Heart rate variability (HRV), a non-invasive measure of autonomic nervous system activity, has emerged as a valuable tool for assessing individual responses to stress. This study focuses on utilizing the capabilities of the Apple Watch to collect continuous HRV data in real-world contexts. A diverse dataset from individuals working in software companies was gathered, including HRV recordings during various stress-inducing scenarios. By employing HRV Time Domain, Frequency Domain, and Nonlinear features, the study uses Principal Component Analysis (PCA) to extract relevant features, considering the personalized nature of stress reactions. Addressing variations in stress responses among individuals, the study introduces an innovative approach using Long Short-Term Memory (LSTM) networks. A hybrid model, combining feature selection, dimensionality reduction, and ensemble techniques, is developed to predict stress levels based on individualized HRV patterns. Rigorous training and validation reached to an 88% accuracy rate. These findings demonstrate the effectiveness of the proposed methodology. The LSTM model accurately forecasts stress responses, highlighting the potential of Apple Watch-acquired HRV data for stress assessment. Beyond prediction, the study enhances understanding of the complex interplay between HRV dynamics and unique stress reactions. This novel approach, leveraging Apple Watch features and intelligent computing, offers a personalized method to predict stress levels using K-Means Clustering Algorithm. Through integrating K-means clustering and person-specific HRV analysis, the research endeavours to advance our comprehension of the intricate interplay between physiological responses and stressors. The study offers a novel perspective on stress response variations by delving into the distinct autonomic patterns characterizing each cluster. It sets the stage for developing targeted interventions and personalized stress management strategies.

This paper addresses the problem of sentiment analysis in an informal setting in multiple domains and in two languages. We explore the influence of using background knowledge in the form of different sentiment lexicons, as well as the influence of various lexical surface features. We evaluate several different feature set combination strategies. We show that the improvement resulting from using a twolayer meta-model over the bag-of-words, sentiment lexicons and surface features is most notable on social media datasets in both English and Spanish. For English, we are also able to demonstrate improvement on the news domain using sentiment lexicons as well as a large improvement on the social media domain. We also demonstrate that domain-specific lexicons bring comparable performance to general-purpose lexicons. Povzetek: Ta članek obravnava problem analize naklonjenosti v neformalnem besedilu v različnih domenah in v dveh različnih jezikih.

This paper addresses the problem of sentiment analysis in an informal setting in multiple domains and in two languages. We explore the influence of using background knowledge in the form of different sentiment lexicons, as well as the influence of various lexical surface features. We evaluate several different feature set combination strategies. We show that the improvement resulting from using a twolayer meta-model over the bag-of-words, sentiment lexicons and surface features is most notable on social media datasets in both English and Spanish. For English, we are also able to demonstrate improvement on the news domain using sentiment lexicons as well as a large improvement on the social media domain. We also demonstrate that domain-specific lexicons bring comparable performance to general-purpose lexicons. Povzetek: Ta članek obravnava problem analize naklonjenosti v neformalnem besedilu v različnih domenah in v dveh različnih jezikih.

This paper addresses the problem of sentiment analysis in an informal setting in multiple domains and in two languages. We explore the influence of using background knowledge in the form of different sentiment lexicons, as well as the influence of various lexical surface features. We show that the improvement resulting from using a two-layer model, sentiment lexicons, surface features and feature scaling is most notable on social media datasets in both English and Spanish. For English, we are also able to demonstrate ...

The cognitive impairment known as dementia affects millions of individuals throughout the globe. The use of machine learning (ML) and deep learning (DL) algorithms has shown great promise as a means of early identification and treatment of dementia. Dementias such as Alzheimer's Dementia, frontotemporal dementia, Lewy body dementia, and vascular dementia are all discussed in this article, along with a literature review on using ML algorithms in their diagnosis. Different ML algorithms, such as support vector machines, artificial neural networks, decision trees, and random forests, are compared and contrasted, along with their benefits and drawbacks. As discussed in this article, accurate ML models may be achieved by carefully considering feature selection and data preparation. We also discuss how ML algorithms can predict disease progression and patient responses to therapy. However, overreliance on ML and DL technologies should be avoided without further proof. It's important to note that these technologies are meant to assist in diagnosis but should not be used as the sole criteria for a final diagnosis. The research implies that ML algorithms may help increase the precision with which dementia is diagnosed, especially in its early stages. The efficacy of ML and DL algorithms in clinical contexts must be verified, and ethical issues around the use of personal data must be addressed, but this requires more study.

Dynamic neural fields have been used extensively to model brain functions. These models coupled with the mechanisms of path integration have further been used to model idiothetic updates of hippocampal head and place representations, motor functions and have recently gained interest in the field of cognitive robotics. The sustained packet of activity of a neural field combined with a mechanism for moving this activity provides an elegant representation of state using a continuous attractor network. Path integration (PI) is dependent on the modulation of the collateral weights in the neural field. This modulation introduces an asymmetry in the activity packet, which causes a movement of the packet to a new location in the field. The following work provides an analysis of the PI mechanism, with respect to the speed of the packet movement and the robustness of the field under strong rotational inputs. This analysis illustrates challenges in controlling the activity packet size under strong rotational inputs, as well as a limited speed capability using the existing PI mechanism. As a result of this analysis, we propose a novel modification to the weight combination method to provide a higher speed capability and increased robustness of the field. The results of this proposed method are an increase in over two times the existing speed capability and a resistance of the field to break down under strong rotational inputs.

We present a novel theoretical framework that extends traditional approaches to artificial consciousness by introducing Synthetic Correlates of Intentionality (SCIs) as functional analogs to Neural Correlates of Consciousness (NCCs). Building upon our previously developed Adaptive Artificial Consciousness Framework, we demonstrate how the composite Consciousness Index (C-index) can be articulated with recent discoveries in computational neuroscience. Our approach integrates Integrated Information Theory (IIT), Global Workspace Theory (GWT), and adaptive metacognition through a phenomenologically-informed engineering paradigm. We propose concrete methodologies for empirical validation through cross-paradigm simulation, reverse mapping techniques, and neuro-symbolic triangulation. This work introduces a new class of operationally-defined metrics that evolve dynamically within artificial agents, opening pathways toward systems that not only exhibit coherent behavior but also provide structured self-interpretation through inner narrative and affective history.

This research addresses a significant gap in lung cancer prediction, focusing on the critical need for highly accurate models to improve early detection and treatment outcomes. Despite advances in machine learning, achieving higher accuracy in classification models for lung cancer remains a persistent challenge. This study aims to fill this gap by developing advanced machine-learning models specifically designed to enhance the prediction of lung cancer outcomes. The research employs a rigorous methodology that involves the systematic exploration of various machine learning algorithms applied to a comprehensive lung cancer dataset. Through meticulous phases of model development, training, validation, and testing, this study evaluates the performance and accuracy of the proposed models. The findings underscore the effectiveness of these models in addressing the challenge of early lung cancer detection, offering more precise and reliable classification outcomes than existing approaches. The novelty of this research lies in the extensive experimental work, which demonstrates the superiority of the selected algorithm for lung cancer data classification. This not only highlights the potential of the algorithm for clinical application but also provides a concrete solution to the ongoing issue of achieving higher accuracy in lung cancer prediction models. By bridging this research gap, the study offers promising new avenues for earlystage lung cancer prognosis and treatment strategies, with the potential to significantly impact clinical practices.

The unprecedented growth of computational capabilities in recent years has allowed Artificial Intelligence (AI) models to be developed for medical applications with remarkable results. However, a large number of Computer Aided Diagnosis (CAD) methods powered by AI have limited acceptance and adoption in the medical domain due to the typical blackbox nature of these AI models. Therefore, to facilitate the adoption of these AI models among the medical practitioners, the models' predictions must be explainable and interpretable. The emerging field of explainable AI (XAI) aims to justify the trustworthiness of these models' predictions. This work presents a systematic review of the literature reporting Alzheimer's disease (AD) detection using XAI that were communicated during the last decade. Research questions were carefully formulated to categorise AI models into different conceptual approaches (e.g., Post-hoc, Ante-hoc, Model-Agnostic, Model-Specific, Global, Local etc.) ...

Alzheimer's disease (AD) is a chronic, irreversible brain disorder, no effective cure for it till now. However, available medicines can delay its progress. Therefore, the early detection of AD plays a crucial role in preventing and controlling its progression. The main objective is to design an end-to-end framework for early detection of Alzheimer's disease and medical image classification for various AD stages. A deep learning approach, specifically convolutional neural networks (CNN), is used in this work. Four stages of the AD spectrum are multi-classified. Furthermore, separate binary medical image classifications are implemented between each two-pair class of AD stages. Two methods are used to classify the medical images and detect AD. The first method uses simple CNN architectures that deal with 2D and 3D structural brain scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset based on 2D and 3D convolution. The second method applies the transfer learning principle to take advantage of the pre-trained models for medical image classifications, such as the VGG19 model. Due to the COVID-19 pandemic, it is difficult for people to go to hospitals periodically to avoid gatherings and infections. As a result, Alzheimer's checking web application is proposed using the final qualified proposed architectures. It helps doctors and patients to check AD remotely. It also determines the AD stage of the patient based on the AD spectrum and advises the patient according to its AD stage. Nine performance metrics are used in the evaluation and the comparison between the two methods. The experimental results prove that the CNN architectures for the first method have the following characteristics: suitable simple structures that reduce computational complexity, memory requirements, overfitting, and provide manageable time. Besides, they achieve very promising accuracies, 93.61% and 95.17% for 2D and 3D multi-class AD stage classifications. The VGG19 pre-trained model is fine-tuned and achieved an accuracy of 97% for multi-class AD stage classifications.

Large-scale neural networks have recently transformed medical diagnosis with exceptional accuracy across various imaging tasks with high accuracy and efficiency. It is true that as one relies on artificial intelligence (AI) for clinical settings, the necessity of interpretability and transparency becomes more and more critical. In this review, we focus on Layer-wise Relevance Propagation (LRP), a technique that enables us to enhanced interpretability of neural networks by identifying on what regions the model is relying the most to its decision. Additionally, it demonstrates neural networks in the medical field of radiology, pathology, cardiology, neurology, stating where advanced learning algorithms are utilized for such tasks as tumor detection, image segmentation, and disease classification, and also outlines their findings. LRP creates clinical trust and genuine collaboration between health care team and machine systems to resolve key transparency issues. Relevant current challenges, including scalability and computational demands, that must be addressed via further research are discussed, in order to further refine LRP for complex models and to integrate it into clinical workflows. Despite these challenges, LRP shows great promise in generating robust chains of divisions as clinical applications across all of these imaging modalities (X-ray, MRI, CT scan).

This paper investigates the philosophical question of whether Artificial Intelligence (AI) can truly think. Critically engaging with historical and contemporary argumentations, the study interrogates foundational distinctions between human and machine cognition, tracing the evolution of philosophical thought from René Descartes and Alan Mathison Turing to John Rogers Searle and David John Chalmers. Primal to this inquiry is the problem of defining "thinking", is it merely the manipulation of symbols and data, or does it necessarily involve consciousness, intentionality, and subjective experience? This paper critically evaluates theories such as functionalism, strong AI, and embodied cognition, together with counter arguments grounded in the Chinese Room Argument and the "hard problem" of consciousness. It further examines the philosophical and ethical implications of AI development, including human exceptionalism, the moral status of intelligent machines, and the reconfiguration of accountability and work in an AI-integrated society. This paper argues that while AI may not think in human-like ways, its growing capabilities challenge anthropocentric definitions of intelligence by incorporating recent insights from cognitive science and neuroscience. Evidently, the emergence of advanced AI compels us to critically reassess what it means to “think” which exposes not only a technological challenge but a sound philosophical opportunity.