To identify objects in images, a complex set of skills is needed that includes understanding the context and being able to determine the borders of objects. In computer vision, this task is known as semantic segmentation and it involves categorizing each pixel in an image. It is crucial in many real-world situations: for autonomous vehicles, it enables the identification of objects in the surrounding area; in medical diagnosis, it enhances the ability to detect dangerous pathologies early, thereby reducing the risk of serious consequences. In this study, we compare the performance of various ensembles of convolutional and transformer neural networks. Ensembles can be created, e.g, by varying the loss function, data augmentation method or the learning rate strategy. Our proposed ensemble, which is based on the simple average rule, demonstrates exceptional performance on several datasets. All the resources used in this study are available online at the following GitHub repository: https://github.com/LorisNanni.

In recent times, BERT based transformer models have become an inseparable part of the 'tech stack' of text processing models. Similar progress is being observed in the speech domain with a multitude of models observing state-of-the-art results by using audio transformer models to encode speech. This begs the question of what are these audio transformer models learning. Moreover, although the standard methodology is to choose the last layer embedding for any downstream task, but is it the optimal choice? We try to answer these questions for the two recent audio transformer models, Mockingjay and wave2vec2.0. We compare them on a comprehensive set of language delivery and structure features including audio, fluency and pronunciation features. Additionally, we probe the audio models' understanding of textual surface, syntax, and semantic features and compare them to BERT. We do this over exhaustive settings for native, non-native, synthetic, read and spontaneous speech datasets

The integration of artificial intelligence (AI) into medical imaging has guided in an era of transformation in healthcare. This literature review explores the latest innovations and applications of AI in the field, highlighting its profound impact on medical diagnosis and patient care. The innovation segment explores into cutting-edge developments in AI, such as deep learning algorithms, convolutional neural networks, and generative adversarial networks, which have significantly improved the accuracy and efficiency of medical image analysis. These innovations have enabled rapid and accurate detection of abnormalities, from identifying tumors during radiological examinations to detecting early signs of eye disease in retinal images. The article also highlights various applications of AI in medical imaging, including radiology, pathology, cardiology, and more. AI-based diagnostic tools not only speed up the interpretation of complex images but also improve early detection of disease, ultimately delivering better outcomes for patients. Additionally, AI-based image processing facilitates personalized treatment plans, thereby optimizing healthcare delivery. This literature review highlights the paradigm shift that AI has brought to medical imaging, highlighting its role in revolutionizing diagnosis and patient care. By combining cutting-edge AI techniques and their practical applications, it is clear that AI will continue shaping the future of healthcare in profound and positive ways.

The usage and amount of information available on the internet increase over the past decade. This digitization leads to the need for automated answering system to extract fruitful information from redundant and transitional knowledge sources. Such systems are designed to cater the most prominent answer from this giant knowledge source to the user’s query using natural language understanding (NLU) and thus eminently depends on the Question-answering(QA) field. Question answering involves but not limited to the steps like mapping of user’s question to pertinent query, retrieval of relevant information, finding the best suitable answer from the retrieved information etc. The current improvement of deep learning models evince compelling performance improvement in all these tasks. In this review work, the research directions of QA field are analyzed based on the type of question, answer type, source of evidence-answer, and modeling approach. This detailing followed by open challenges of the field like automatic question generation, similarity detection and, low resource availability for a language. In the end, a survey of available datasets and evaluation measures is presented.

Anomaly detection and segmentation aim at distinguishing abnormal images from normal images and further localizing the anomalous regions. Feature reconstruction based method has become one of the mainstream methods for this task. This kind of method has two assumptions: (1) The features extracted by neural network is a good representation of the image. (2) The autoencoder solely trained on the features of normal images cannot reconstruct the features of anomalous regions well. But these two assumptions are hard to meet. In this paper, we propose a new anomaly segmentation method based on feature reconstruction. Our approach mainly consists of two parts: (1) We use a pretrained vision transformer (ViT) to extract the features of the input image. (2) We design a self-attention autoencoder to reconstruct the features. We regard that the self-attention operation which has a global receptive field is beneficial to the methods based on feature reconstruction both in feature extraction and reconstruction. The experiments show that our method outperforms the state-of-the-art approaches for anomaly segmentation on the MVTec dataset. It is both effective and time-efficient.

Image classification in Remote Sensing and Geographic Information Systems (GIS) containing various land-cover classes is essential for efficient and sustainable land-use estimation, and other tasks like object detection, localization and segmentation. Deep Learning (DL) techniques have shown a tremendous potential in the GIS domain. While Convolutional Neural Networks (CNN) have dominated most of the image analysis domain, a new architecture called transformers have proved to be a unifying solution for several AI-based processing pipelines. Vision Transformers (ViT), a variant of transformers can have comparable and in some cases better accuracy than a CNN. However, they suffer from a significant drawback associated with an excessive use of training parameters. In this research we explore several modifications in the vision transformer architectures, especially MobileViT that can be optimized while boosting accuracy. To verify our proposed approach these new architectures are trained on four land-cover datasets AID, EuroSAT, UC-Merced, and WHU-RS19. Experiments reveal that combination of lightweight convolutional layers including ShuffleNet along with depthwise separable convolutions and average pooling can reduce the trainable parameters by 17.85% and yet achieve higher accuracy than the base MobileViT. It is also observed that utilizing a combination of convolution layers along with multi-headed self attention layers in MobileViT variants provide better performance in capturing local and global features unlike the standalone ViT architecture that utilizes almost 95% more parameters than the proposed MobileViT variant.

Real-time tracking is one of the most challenging problems in computer vision. Most Transformer-based trackers usually require expensive computational and storage power, which leads to the inability of these robust trackers to achieve real-time performance in resource-constrained devices. In this work, we propose a lightweight tracker, AnteaTrack. To localize the target more precisely, we develop a scaling-invariant max-filtering operator employing a sliding window combined with local max-pooling, which filters out the suspected target from the feature and performs an augmented representation while suppressing the background. For a more tight bounding box, we employ Pixel-Shuffle to increase the resolution of the feature map and get a more fine-grained representation of the target. In addition, AnteaTrack can run in real-time at 47 frames per second(FPS) on the CPU. We tested AnteaTrack on 5 datasets, and a large number of experiments have shown that AnteaTrack provides the most efficient solution compared to the same type of CPU real-time trackers. The code will be available at https://github.com/cnchange/AnteaTrack.

In the realm of computer vision, semantic segmentation is the task of recognizing objects in images at the pixel level. This is done by performing a classification of each pixel. The task is complex and requires sophisticated skills and knowledge about the context to identify objects’ boundaries. The importance of semantic segmentation in many domains is undisputed. In medical diagnostics, it simplifies the early detection of pathologies, thus mitigating the possible consequences. In this work, we provide a review of the literature on deep ensemble learning models for polyp segmentation and we develop new ensembles based on convolutional neural networks and transformers. The development of an effective ensemble entails ensuring diversity between its components. To this end, we combine different models (HarDNet-MSEG, Polyp-PVT, and HSNet) trained with different data augmentation techniques, optimization methods, and learning rates, which we experimentally demonstrate to be useful to form a better ensemble. Most importantly, we introduce a new method to obtain the segmentation mask which is more suitable for combining transformers in an ensemble. In our extensive experimental evaluation, the proposed ensembles exhibit state-of-the-art performance.

Text summarization remains a challenging task in the Natural Language Processing field despite the plethora of applications in enterprises and daily life. One of the common use cases is the summarization of web pages which has the potential to provide an overview of web pages to devices with limited features. In fact, despite the increasing penetration rate of mobile devices in rural areas, the bulk of those devices offer limited features in addition to the fact that these areas are covered with limited connectivity such as the GSM network. Summarizing web pages into SMS becomes, therefore, an important task to provide information to limited devices. This work introduces WATS-SMS, a T5-based French Wikipedia Abstractive Text Summarizer for SMS. It is built through a transfer learning approach. The T5 English pre-trained model is used to generate a French text summarization model by retraining the model on 25,000 Wikipedia pages then compared with different approaches in the literature. The objective is twofold: (1) to check the assumption made in the literature that abstractive models provide better results compared to extractive ones; and (2) to evaluate the performance of our model compared to other existing abstractive models. A score based on ROUGE metrics gave us a value of 52% for articles with length up to 500 characters against 34.2% for transformer-ED and 12.7% for seq-2seq-attention; and a value of 77% for articles with larger size against 37% for transformers-DMCA. Moreover, an architecture including a software SMS-gateway has been developed to allow owners of mobile devices with limited features to send requests and to receive summaries through the GSM network.

Since large power transformers are custom-made, and their design process is a labor-intensive task, their design process is split into different parts. In tendering, the price calculation is based on the preliminary design of the transformer. Due to the complexity of this task, it belongs to the most general branch of discrete, non-linear mathematical optimization problems. Most of the published algorithms are using a copper filling factor based winding model to calculate the main dimensions of the transformer during this first, preliminary design step. Therefore, these cost optimization methods are not considering the detailed winding layout and the conductor dimensions. However, the knowledge of the exact conductor dimensions is essential to calculate the thermal behaviour of the windings and make a more accurate stray loss calculation. The paper presents a novel, evolutionary algorithm-based transformer optimization method which can determine the optimal conductor shape for the windings during this examined preliminary design stage. The accuracy of the presented FEM method was tested on an existing transformer design. Then the results of the proposed optimization method have been compared with a validated transformer design optimization algorithm.

In the realm of digital libraries, efficiently managing and accessing scientific publications necessitates automated bibliographic reference segmentation. This study addresses the challenge of accurately segmenting bibliographic references, a task complicated by the varied formats and styles of references. Focusing on the empirical evaluation of Conditional Random Fields (CRF), Bidirectional Long Short-Term Memory with CRF (BiLSTM+CRF), and Transformer Encoder with CRF (Transformer+CRF) architectures, this research employs Byte Pair Encoding and Character Embeddings for vector representation. The models underwent training on the extensive Giant corpus and subsequent evaluation on the Cora Corpus to ensure a balanced and rigorous comparison, maintaining uniformity across embedding layers, normalization techniques, and Dropout strategies. Results indicate that the BiLSTM+CRF architecture outperforms its counterparts by adeptly handling the syntactic structures prevalent in bibliographic data, achieving an F1-Score of 0.96. This outcome highlights the necessity of aligning model architecture with the specific syntactic demands of bibliographic reference segmentation tasks. Consequently, the study establishes the BiLSTM+CRF model as a superior approach within the current state-of-the-art, offering a robust solution for the challenges faced in digital library management and scholarly communication.

In this review paper, we discuss the use of BERT, one of the most well-liked deep learning-based language models. The model's operation mechanism, key areas of applicability to text analytics tasks, comparisons with related models for each activity, and a description of certain proprietary models are all covered in this work. The data from several dozen original scientific studies that have been published in the last few years and have garnered the greatest interest from the scientific community were systematized in order to prepare this review. All researchers and students who wish to learn about the most recent developments in the field of natural language text analysis may find this survey helpful. Bidirectional Encoder Representations from Transformers, or BERT for short, is the subject of a thorough investigation that we present. The area of natural language processing holds significant importance in the creation of intelligent systems. To do a variety of activities, one must comprehend the sentence's correct meaning in order to produce the desired result. Computers have a hard time understanding languages because of how context constantly changes. The main challenge in natural language processing tasks is getting computers to understand text context, and BERT is seen as a revolution in this regard. It picks up the language and its meaning in a manner that is quite similar to how the human brain processes meaning from sentences. Its capacity to identify a word from both the left and right contexts of a sentence makes it special. A new era in the perception and comprehension of natural languages has been ushered in with the development of BERT, which could help computers understand natural languages more fully. This study aims to provide readers with a deeper knowledge of the BERT language model and how it is applied to different NLP tasks.

A key component for the proper functioning, availability, and reliability of power grids is the power trans- former. Although these are very reliable assets, the early detection of incipient degradation mechanisms are very important to prevent failures that may shorten their lifetime. In this work a review and comparative analysis, of classical Machine Learning algorithms (such as single and ensemble classification algorithms) and two automatic machine learning classifiers, are presented for the fault diagnosis of power transformers. The goal is to determine whether fully automated ML approaches are worse or better than traditional ML frameworks that require a human in the loop (such as a data scientist) to identify transformer faults from oil-dissolved gases data. The methodology uses a DB compiled from published and proprietary transformer fault samples. Faults data is obtained from the literature, the Duval pentagon method, and user-expert knowledge. The parameters from, either single or ensemble classifiers, were optimized through standard machine learning procedures. The results showed that the best performing algorithm is a robust automatic machine learning classifiers model followed by classical algorithms such as neural networks and stacking ensembles. These results highlight the ability of a robust automatic machine learning model for, handling imbalanced power transformers faults datasets with high accuracy, using the minimum tuning effort by the electrical experts. By providing the most probable transformer fault condition will reduce the time required to find and solve the fault.

This paper applies a new artificial intelligence architecture, the Temporal Fusion Transformer (TFT), for the joint GDP forecasting of 25 OECD countries at different time horizons. This new attention-based architecture offers significant advantages over other deep learning methods. First, results are interpretable since the impact of each explanatory variable on each forecast can be calculated. Second, it allows to visualize persistent temporal patterns and to identify significant events and different regimes. Third, it provides quantile regressions and permits to train the model on multiple time series from different distributions. Results suggest that TFTs outperform regression models, especially in periods of turbulence such as the COVID-19 shock. Interesting economic interpretations are obtained depending on whether the country is domestic demand-led or export-led growth. In essence, TFT is revealed as a new tool that artificial intelligence provides to economists and policy makers, with enormous prospects for the future.

In this paper we propose a novel self-supervised approach of keywords and keyphrases retrieval and extraction by an end-to-end deep learning approach, which is trained by contextually self-labelled corpus. Our proposed approach is novel to use contextual and semantic features to extract the keywords and has outperformed the state of the art. Through the experiment the proposed approach has been proved to be better in both semantic meaning and quality than the existing popular algorithms of keyword extraction. In addition, we propose to use contextual features from bidirectional transformers to automatically label short-sentence corpus with keywords and keyphrases to build the ground truth. This process avoids the human time to label the keywords and do not need any prior knowledge. To the best of our knowledge, our published dataset in this paper is a fine domain-independent corpus of short sentences with labelled keywords and keyphrases in the NLP community.

The predominance of English and Latin-based large language models (LLMs) has led to a notable deficit in native Arabic LLMs. This discrepancy is accentuated by the prevalent inclusion of English tokens in existing Arabic models, detracting from their efficacy in processing native Arabic's intricate morphology and syntax. Consequently, there is a theoretical and practical imperative for developing LLMs predominantly focused on Arabic linguistic elements. To address this gap, this paper proposes ArabianGPT, a series of transformer-based models within the ArabianLLM suite designed explicitly for Arabic. These models, including ArabianGPT-0.1B and ArabianGPT-0.3B, vary in size and complexity, aligning with the nuanced linguistic characteristics of Arabic. The AraNizer tokenizer, integral to these models, addresses the unique morphological aspects of Arabic script, ensuring more accurate text processing. Empirical results from fine-tuning the models on tasks like sentiment analysis and summarization demonstrate significant improvements. For sentiment analysis, the fine-tuned ArabianGPT-0.1B model achieved a remarkable accuracy of 95%, a substantial increase from the base model's 56%. Similarly, in summarization tasks, fine-tuned models showed enhanced F1 scores, indicating improved precision and recall in generating concise summaries. Comparative analysis of fine-tuned ArabianGPT models against their base versions across various benchmarks reveals nuanced differences in performance, with fine-tuning positively impacting specific tasks like question answering and summarization. These findings underscore the efficacy of fine-tuning in aligning ArabianGPT models more closely with specific NLP tasks, highlighting the potential of tailored transformer architectures in advancing Arabic NLP.

When integrating data from different sources, there are problems of synonymy, different languages, concepts of different granularity. This paper proposes a simple but effective approach to evaluate the semantic similarity of short texts, especially keywords. The method is capable of matching keywords from different sources and languages by exploiting transformers and WordNet-based methods. Key features of the approach include its unsupervised pipeline, mitigation of the lack of context in keywords, scalability for large archives, support for multiple languages and real-world scenarios adaptation capabilities. The work aims to provide a versatile tool for different cultural heritage archives without requiring complex customization. The objectives of the paper are to explore different approaches to identifying similarities in 1- or n-gram tags, to evaluate and compare different pre-trained language models, and to define integrated methods to overcome limitations. Tests to validate the approach have been conducted using the QueryLab portal, a search engine for cultural heritage archives, to evaluate the proposed pipeline.

Unmanned Aerial Vehicles (UAV) are increasingly being used in a variety of domains and precision agriculture is no exception. Precision agriculture is the future of agriculture and will play a key role in long-term sustainability of agricultural practices. This paper presents a survey of how image data collected using UAVs has been used in conjunction with ma-chine learning techniques to support precision agriculture. Numerous agricultural applications including classification of crop types and trees, crops detection, weed detection, cropland cover, and segmentation of farming fields are discussed. A variety of supervised, semi-supervised and unsupervised machine learning techniques for image-based preci-sion agriculture are compared. The survey showed that for traditional machine learning approaches, Random Forests performed better than Support Vector Machines (SVM) and K-Nearest Neighbor Algorithm (KNN) for crop/weed classification. And, while Convolutional Neural Networks (CNN) have been used extensively, the U-Net-based models out-performed conventional CNN models for classification and segmentation tasks. Among the Single Stage Detectors (SSD), YOLO series performed relatively well. Two-Stage Detectors like R-CNN, FPN, and Mask R-CNN generally tended to outperform SSDs. Vision Trans-formers (ViT) showed promising results amongst transformer-based models which did not generally perform better than CNNs. Finally, Generative Adversarial Networks (GANs) have been used to address the problem of smaller datasets and unbalanced data

Sea urchins are long-living invertebrates with a complex immune system which includes extended families of immune receptors. A central immune gene family in the sea urchins encodes for the Transformer (Trf) proteins. The Trf family was so far studied mainly in the purple sea urchin Strongylocentrotus purpuratus. In this study, we explored this protein family in the Mediterranean Sea urchin Paracentrotus lividus. The PlTrf genes and predicted proteins were found to be highly diverse and showed a typical Trf size range and structure. We found that P. lividus coelomocytes and hemolymph contain different PlTrf protein repertoires with a shared subset which specifically bind E. coli bacteria. Using FACS, we identified five different P. lividus coelomocyte sub-populations with cell surface Trf protein expression. The relative abundance of the Trf-positive cells sharply increased following immune challenge with E. coli bacteria, but not following challenge with LPS or sea urchin pathogen V. penaeicida. Finally, we demonstrated that the phagocytosis of E. coli bacteria by P. lividus phagocytes is mediated through the hemolymph and is inhibited by blocking Trf activity with anti-Trf antibodies. Together, our results suggest collaboration between cellular and humoral Trf-mediated effector arms in the P. lividus specific immune response to pathogens.

Blood cell analysis is a crucial diagnostic process in medical practice. In particular, detecting white blood cells (WBCs) is essential for diagnosing of many diseases. The manual screening of blood films is a time-consuming and subjective process, which can lead to inconsistencies and errors. Therefore, automated detection of blood cells can improve the accuracy and efficiency of the screening process. In this study, an explainable Vision Transformer (ViT) model was proposed for the automatic detection of WBCs from blood films. The proposed model utilizes the self-attention mechanism to extract relevant features from the input images and leverages transfer learning by incorporating pre-trained model weights to improve its performance. The proposed model achieved a classification accuracy of 99.40% for five distinct types of WBCs and exhibited potential in reducing the time required for manual screening of blood films by pathologists. Upon examination of the misclassified test samples, it was observed that incorrect predictions were correlated with the presence or absence of granules in the cell samples. To validate this observation, the dataset was divided into two classes, namely Granulocytes and Agranulocytes, and a secondary training process was conducted. The resulting ViT model trained for binary classification achieved an accuracy of 99.70%, recall of 99.54%, precision of 99.32%, and F-1 score of 99.43% during the test phase. To ensure the reliability of the ViT model's multi-class classification of WBCs, the pixel areas that the model focuses on in its predictions are visualized through the Score-CAM algorithm.

There is great interest in automatically detecting road weather and understanding its impacts on the overall safety of the transport network. This can, for example, support road condition-based maintenance or even serve as detection systems that assist safe driving during adverse climate conditions. In computer vision, previous work has demonstrated the effectiveness of deep learning in predicting weather conditions from outdoor images. However, training deep learning models to accurately predict weather conditions using real-world road-facing images is difficult due to: (1) the simultaneous occurrence of multiple weather conditions; (2) imbalanced occurrence of weather conditions throughout the year; and (3) road idiosyncrasies, such as road layouts, illumination, road objects etc. In this paper, we explore the use of focal loss function to force the learning process to focus on weather instances that are hard to learn with the objective to help address data imbalance. In addition, we explore the attention mechanism for pixel based dynamic weight adjustment to handle road idiosyncrasies using state-of-the-art vision transformer models. Experiments with a novel multi-label road weather dataset show that focal loss significantly increases the accuracy of computer vision approaches for imbalanced weather conditions. Furthermore, vision transformers outperforms current state-of-the-art convolutional neural networks in predicting weather conditions with a validation accuracy of 92% and F1-score of 81.22%, which is impressive considering the imbalanced nature of the dataset.

The classification of goods involved in international trade in Brazil is based on the Mercosur Common Nomenclature (NCM). The classification of these goods represents a real challenge due to the complexity involved in assigning the correct category codes especially considering the legal and fiscal implications of misclassification. This work focuses on the training of a classifier based on Bidirectional En-coder Representations from Transformers (BERT) for the tax classification of goods with NCM codes. In particular, this article presents results from using a specific Portuguese Language tuned BERT model as well results from using a Multilingual BERT. Experimental results justify the use of these models in the classification process and also that the language specific model has a slightly better performance.

This study leverages transformer-based models and focuses on data representation strategies in the named entity recognition task, including "single" (one sentence per vector), "merged" (multiple sentences per vector), and "context" (sentences joined with attention to context). Performance analysis reveals that models trained with a single strategy may not perform well on different data representations. A combined training procedure is proposed to address this limitation, using all three strategies to enhance the stability and adaptability of the model. The results of this approach are presented and discussed for various datasets for four languages (English, Polish, Czech, and German), demonstrating the effectiveness of the combined strategy.

The ‘inverse problem’ of mass spectrometric molecular identification (‘given a mass spectrum, calculate/predict the 2D structure of the molecule whence it came’) is largely unsolved, and is especially acute in metabolomics where many small molecules remain unidentified. This is largely because the number of experimentally available electrospray mass spectra of small molecules is quite limited. However, the forward problem (‘calculate a small molecule’s likely fragmentation and hence at least some of its mass spectrum from its structure alone’) is much more tractable, because the strengths of different chemical bonds are roughly known. This kind of molecular identification problem may be cast as a language translation problem in which the source language is a list of high-resolution mass spectral peaks and the ‘translation’ a representation (for instance in SMILES) of the molecule. It is thus suitable for attack using the deep neural networks known as transformers. We here present MassGenie, a method that uses a transformer-based deep neural network, trained on ~6 million chemical structures with augmented SMILES encoding and their paired molecular fragments as generated in silico, explicitly including the protonated molecular ion. This architecture (containing some 400 million elements) is used to predict the structure of a molecule from the various fragments that may be expected to be observed when some of its bonds are broken. Despite being given essentially no detailed nor explicit rules about molecular fragmentation methods, isotope patterns, rearrangements, neutral losses, and the like, MassGenie learns the effective properties of the mass spectral fragment and valency space, and can generate candidate molecular structures that are very close or identical to those of the ‘true’ molecules. We also use VAE-Sim, a previously published variational autoencoder, to generate candidate molecules that are ‘similar’ to the top hit. In addition to using the ‘top hits’ directly, we can produce a rank order of these by ‘round-tripping’ candidate molecules and comparing them with the true molecules, where known. As a proof of principle, we confine ourselves to positive electrospray mass spectra from molecules with a molecular mass of 500Da or lower, including those in the last CASMI challenge (for which the results are known), getting 49/93 (53%) precisely correct. The transformer method, applied here for the first time to mass spectral interpretation, works extremely effectively both for mass spectra generated in silico and on experimentally obtained mass spectra from pure compounds. It seems to act as a Las Vegas algorithm, in that it either gives the correct answer or simply states that it cannot find one. The ability to create and to ‘learn’ millions of fragmentation patterns in silico, and therefrom generate candidate structures (that do not have to be in existing libraries) directly, thus opens up entirely the field of de novo small molecule structure prediction from experimental mass spectra.

The question of molecular similarity is core in cheminformatics, and is usually assessed via a pairwise comparison based on vectors of properties or molecular fingerprints. We recently exploited variational autoencoders to embed 6M molecules in a chemical space, such that their (Euclidean) distance within the latent space so formed could be assessed within the framework of the entire molecular set. However, the standard objective function used did not seek to manipulate the latent space so as to cluster the molecules based on any perceived similarity. Using a set of some 160,000 molecules of biological relevance, we here bring together three modern elements of deep learning to create a novel and disentangled latent space, viz transformers, contrastive learning, and an embedded autoencoder. The effective dimensionality of the latent space was varied such that clear separation of individual types of molecules could be observed within individual dimensions of the latent space. The capacity of the network was such that many dimensions were not populated at all. As before, we assessed the utility of the representation by comparing clozapine with its near neighbours, and did the same for various antibiotics related to flucloxacillin. Transformers, especially when as here coupled with contrastive learning, effectively provide one-shot learning, and lead to a successful and disentangled representation of molecular latent spaces that at once uses the entire training set in their construction while allowing ‘similar’ molecules to cluster together in an effective and interpretable way.

Social media has become more popular in the last few years. It has been used in public health development and healthcare settings to promote healthier lifestyles. Given its important role in today's culture, it is necessary to understand its current trends and future directions in public health. This review aims to describe and summarize how public health professionals have been using social media to improve population outcomes. Additionally, it discusses potential future directions on how social media can be used to improve population health. The findings showed how social media has been used as a tool for research, designing and implementing health campaigns, messaging, and health promotion. In addition, social media integration with Artificial Intelligence (AI) and Generative Pre-Trained transformers (GPT) can have a big impact and offer an innovative approach to tackle the problems and difficulties in health informatics. These innovations might improve patient participation, individualized assistance, and care experiences. The research showed how likely social media will keep growing and evolving and, if used effectively, have the potential to help close public health gaps across different cultures and improve population health.

The disruptive application of ChatGPT (GPT-3.5, GPT-4) to a variety of domains has become a topic of much discussion in the scientific community and society at large. Large Language Models (LLMs), e.g., BERT, Bard, Generative Pre-trained Transformers (GPTs), LLaMA, etc., have the ability to take instructions, or prompts, from users and generate answers and solutions based on very large volumes of text-based training data. This paper assesses the impact and potential impact of ChatGPT on the field of digital forensics, specifically looking at its latest pre-trained LLM, GPT-4. A series of experiments are conducted to assess its capability across several digital forensic use cases including artefact understanding, evidence searching, code generation, anomaly detection, incident response, and education. Across these topics, its strengths and risks are outlined and a number of general conclusions are drawn. Overall this paper concludes that while there are some potential low-risk applications of ChatGPT within digital forensics, many are either unsuitable at present, since the evidence would need to be uploaded to the service, or they require sufficient knowledge of the topic being asked of the tool to identify incorrect assumptions, inaccuracies, and mistakes. However, to an appropriately knowledgeable user, it could act as a useful supporting tool in some circumstances.

The aim of this document is to present how the interpretation of the RVM (Recovery Voltage Measurement) test can be improved through the use of a Debye equivalent circuit. As it is described in the literature the interpretation of the RVM test requires expertise and if the transformer presents a high interfacial polarization it is not possible to diagnose it in detail. Debye model is proposed in this work for enhancing RVM interpretation. This model is based on an electrical circuit that includes basic R-L-C components, that allows two interesting features: on one hand, isolation physical effects can be separately represented and, on the other, the values of the R-C components can be calculated from the RVM response (L components are not used in this work as long as no magnetic field effects are taken into account). Thus, the different isolation effects that are indistinguishable merged in the RVM transient response can be split into different R-C branches, each one corresponding with a single (not merged) isolation effect. Finally, several case studies are presented, in which it is correlated a dielectric oil treatment carried out and the equivalent circuit changes.

The central goal of few-shot scene classification is to learn a model that can generalize well to a novel scene category (UNSEEN) from only one or a few labeled examples. Recent works in the remote sensing (RS) community tackle this challenge by developing algorithms in a meta-learning manner. However, most prior approaches have either focused on rapidly optimizing a meta-learner or aimed at finding good similarity metrics while overlooking the embedding power. Here we propose a novel Task-Adaptive Embedding Learning (TAEL) framework that complements the existing methods by giving full play to feature embedding’s dual roles in few-shot scene classification - representing images and constructing classifiers in the embedding space. First, we design a lightweight network that enriches the diversity and expressive capacity of embeddings by dynamically fusing information from multiple kernels. Second, we present a task-adaptive strategy that helps to generate more discriminative representations by transforming the universal embeddings into task-specific embeddings via a self-attention mechanism. We evaluate our model in the standard few-shot learning setting on two challenging datasets: NWPU-RESISC4 and RSD46-WHU. Experimental results demonstrate that, on all tasks, our method achieves state-of-the-art performance by a significant margin.

Cooling fans are one of the critical components of air-forced (AF) dry-type transformers for regulating internal temperatures. Therefore, effective malfunction detection is crucial to maintain the transformer temperature within an acceptable range and prevent overheating. Regular maintenance occurs periodically and issues with cooling fans may arise between maintenance periods, leading to prolonged operation under malfunctioning conditions and potential failures. In addition, utilities typically have online information about whether a fan works or not without providing information about cooling fan malfunctioning circumstances. To address these challenges, this study proposes learning-based online monitoring techniques to detect malfunctions in AF transformer cooling fans. Random forests (RFs) and convolutional neural networks (CNNs) are developed to classify the audio signals from cooling fans into normal and malfunctioning classes. Unlike RFs, which require separate feature extraction, CNNs are trained based on spectrogram images derived from audio signals. Thus, various time-frequency techniques are utilized for feature extraction in RFs. Besides, multiple data augmentation techniques are employed to enhance the dataset size and diversity. Algorithmic performance is optimized through hyperparameter tuning and classifier threshold adjustment. Simulations reveal that CNNs outperform RFs, whereas the latter provides superior interpretability of acoustic features compared to the former.

(1) Background: Artificial Intelligence (AI) models have shown potential in various educational contexts. However, their utility in explaining complex biological phenomena, such as Intrinsically Disordered Proteins (IDPs), requires further exploration. This study empirically evaluated the performance of various Large Language Models (LLMs) in the educational domain of IDPs. (2) Methods: Four LLMs, GPT-3.5, GPT-4, GPT-4 with Browsing, and Google Bard (PaLM 2), were assessed using a set of IDP-related questions. An expert evaluated their responses across five categories: accuracy, relevance, depth of understanding, clarity, and overall quality. Descriptive statistics, ANOVA, and Tukey's honesty significant difference tests were utilized for analysis. (3) Results: The GPT-4 model consistently outperformed the others across all evaluation categories. Although GPT-4 and GPT-3.5 were not statistically significantly different in performance (p>0.05), GPT-4 was preferred as the best response in 13 out of 15 instances. The AI models with browsing capabilities, GPT-4 with Browsing and Google Bard (PaLM 2) displayed lower performance metrics across the board with statistically significant differences (p<0.0001). (4) Conclusion: Our findings underscore the potential of AI models, particularly LLMs such as GPT-4, in enhancing scientific education, especially in complex domains such as IDPs. Continued innovation and collaboration among AI developers, educators, and researchers are essential to fully harness the potential of AI for enriching scientific education.

Oil spillage over a sea or ocean’s surface is a threat to marine and coastal ecosystems. Spaceborne synthetic aperture radar (SAR) data has been used efficiently for the detection of oil spills due to its operational capability in all-day all-weather conditions. The problem is often modeled as a semantic segmentation task. The images need to be segmented into multiple regions of interest such as sea surface, oil spill, look-alikes, ships and land. Training of a classifier for this task is particularly challenging since there is an inherent class imbalance. In this work, we train a convolutional neural network (CNN) with multiple feature extractors for pixel-wise classification; and introduce to use a new loss function, namely ‘gradient profile’ (GP) loss, which is in fact the constituent of the more generic Spatial Profile loss proposed for image translation problems. For the purpose of training, testing and performance evaluation, we use a publicly available dataset with selected oil spill events verified by the European Maritime Safety Agency (EMSA). The results obtained show that the proposed CNN trained with a combination of GP, Jaccard and focal loss functions can detect oil spills with an intersection over union (IoU) value of 63.95%. The IoU value for sea surface, look-alikes, ships and land class is 96.00%, 60.87%, 74.61% and 96.80%, respectively. The mean intersection over union (mIoU) value for all the classes is 78.45%, which accounts for a 13% improvement over the state of the art for this dataset. Moreover, we provide extensive ablation on different Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) based hybrid models to demonstrate the effectiveness of adding GP loss as an additional loss function for training. Results show that GP loss significantly improves the mIoU and F1 scores for CNNs as well as ViTs based hybrid models. GP loss turns out to be a promising loss function in the context of deep learning with SAR images.

The Global natural and manmade events are exposing the fragility of the tourism industry and its impact on the global economy. Prior to the COVID-19 pandemic, tourism contributed 10.3% to the global GDP and employed 333 million people but saw a significant decline due to the pandemic. Sustainable and smart tourism requires collaboration from all stakeholders and a comprehensive understanding of global and local issues to drive responsible and innovative growth in the sector. This paper presents an approach for leveraging big data and deep learning to dis-cover holistic, multi-perspective (e.g., local, cultural, national, and international) and objective information on a subject. Specifically, we develop a machine learning pipeline to extract parameters from academic literature and public opinions on Twitter, providing a unique and comprehensive view of the industry from both academic and public perspectives. The academic-view dataset was created from the Scopus database and contains 156,759 research articles from 2000 to 2022, which were modelled to identify 33 distinct parameters in 4 categories: Tourism Types, Planning, Challenges, and Media & Technologies. A Twitter dataset of 485,813 tweets was collected over 18 months starting March 2021 to August 2022 to showcase public perception of tourism in Saudi Arabia, which was modelled to reveal 13 parameters categorized into two broader sets: Tourist Attractions and Tourism Services. Discovering system parameters are re-quired to embed autonomous capabilities in systems and for decision-making and problem-solving during system design and operations. The proposed approach improves AI-based information discovery by extending the use of scientific literature, Twitter, and other sources for autonomous, dynamic optimizations of systems, promoting novel research in the tourism sector and contributing to the development of smart and sustainable societies. The paper also presents a comprehensive knowledge structure and literature review of the tourism sector based on over 250 research articles.

Technological advancements and innovations have profoundly changed the lives of people giving rise to smart environments, cities, and societies. As homes are the building block of cities and societies, smart homes are critical to establishing smart living and are expected to play a key role in enabling smart cities and societies. The current academic literature and commercial advancements on smart homes have mainly focused on developing and providing smart functions for homes to provide security management and facilitate the residents in their various activities such as ambiance management. Homes are much more than physical structures, buildings, appliances, operational machines, and systems. Homes are composed of families and are inherently complex phenomena underlined by humans and their relationships with each other, subject to individual, intragroup, intergroup, and intercommunity goals. There is a clear need to understand, define, consolidate existing research, and actualize the overarching roles of smart homes, the roles of smart homes that would serve the needs of future smart cities and societies. This paper introduces our data-driven parameter discovery methodology and uses it to provide, for the first time, an extensive, rather fairly comprehensive, analysis of the families and homes landscape seen through the eyes of academics and the public using over a hundred thousand research papers and nearly a million tweets. We develop a methodology using deep learning, natural language processing (NLP), and big data analytics methods and apply it to automatically discover parameters that capture a comprehensive knowledge and design space of smart families and homes comprising social, political, economic, environmental, and other dimensions. The 66 discovered parameters and the knowledge space comprising 100s of dimensions are explained by reviewing and referencing over 300 articles from the academic literature and tweets. The knowledge and parameters discovered in this paper can be used to develop a holistic understanding of matters related to families and homes facilitating the development of better, community-specific, policies, technologies, solutions, and industries for families and homes, leading to strengthening families and homes, and in turn, empowering sustainable societies across the globe.

Predatory research field in science is when an emerging scientific topic destroys current topics and characterizes a main scientific change. Predatory research field can be a basic driver of scientific and technological change that generates a 'creative destruction' in science and society in contexts of knowledge-based competition and rapid changes. The prediction of proposed theory of predatory research fields is that it destroys with a fast growth other research fields. The theoretical approach is tested here in research fields of large language models (LLM) by analyzing the transformers (a deep learning architecture based on the multi-head attention mechanism) proposed in 2017 and from November 2022 started main applications in generative artificial intelligence with innovations of BERT, ChatGPT, Microsoft Copilot (launched on February 7, 2023) and other natural language processing tools driven by AI technology for engaging conversations, gain insights, automate tasks, etc., etc. Statistical evidence suggests that growth rate of transformer technologies is 80.58%, a high level compared to all other research fields in machine learning (having a growth rate of 13.83%). Moreover, predatory research field of transformers has a destructive power such that all other domains in LLM from 2021 to 2023 have a general reduction of scientific growth. The impact of transformers is much more drastic of previous radical technologies such as CNN having a temporal growth rate of 0.16%, lower than 0.38% by transformers, ceteris paribus. These analysis reveals that transformers have characteristics to generate a radical scientific and technological change in a not-too-distant future. Overall, then, the study suggests that predatory research field on emerging topics and technologies can generate path-breaking innovations and the examination here can clarify the essential elements of the science dynamics for a better theory of scientific and technological change, providing also main implications for knowledge policy to support promising research fields and technologies to guide economic and social change.