Search

Transformers have revolutionized almost all natural language processing (NLP) tasks but suffer from memory and computational complexity that scales quadratically with sequence length. In contrast, recurrent neural networks (RNNs) exhibit linear scaling in memory and computational requirements but struggle to match the same performance as Transformers due to limitations in parallelization and scalability. We propose a novel model architecture, Receptance Weighted Key Value (RWKV), that combines the efficient parallelizable training of Transformers with the efficient inference of RNNs. Our approach leverages a linear attention mechanism and allows us to formulate the model as either a Transformer or an RNN, which parallelizes computations during training and maintains constant computational and memory complexity during inference, leading to the first non-transformer architecture to be scaled to tens of billions of parameters. Our experiments reveal that RWKV performs on par with similarly sized Transformers, suggesting that future work can leverage this architecture to create more efficient models. This work presents a significant step towards reconciling the trade-offs between computational efficiency and model performance in sequence processing tasks.

OpenAI has released the Chat Generative Pre-trained Transformer (ChatGPT) and revolutionized the approach in artificial intelligence to human-model interaction. Several publications on ChatGPT evaluation test its effectiveness on well-known natural language processing (NLP) tasks. However, the existing studies are mostly non-automated and tested on a very limited scale. In this work, we examined ChatGPT's capabilities on 25 diverse analytical NLP tasks, most of them subjective even to humans, such as sentiment analysis, emotion recognition, offensiveness, and stance detection. In contrast, the other tasks require more objective reasoning like word sense disambiguation, linguistic acceptability, and question answering. We also evaluated GPT-4 model on five selected subsets of NLP tasks. We automated ChatGPT and GPT-4 prompting process and analyzed more than 49k responses. Our comparison of its results with available State-of-the-Art (SOTA) solutions showed that the average loss in quality of the ChatGPT model was about 25% for zero-shot and few-shot evaluation. For GPT-4 model, a loss for semantic tasks is significantly lower than for ChatGPT. We showed that the more difficult the task (lower SOTA performance), the higher the ChatGPT loss. It especially refers to pragmatic NLP problems like emotion recognition. We also tested the ability to personalize ChatGPT responses for selected subjective tasks via Random Contextual Few-Shot Personalization, and we obtained significantly better user-based predictions. Additional qualitative analysis revealed a ChatGPT bias, most likely due to the rules imposed on human trainers by OpenAI. Our results provide the basis for a fundamental discussion of whether the high quality of recent predictive NLP models can indicate a tool's usefulness to society and how the learning and validation procedures for such systems should be established., Comment: preprint

Sistemas de detecção de intrusões modernos são comumente desenvolvidos com uso de algoritmos de aprendizado de máquina e seleção de atributos. No entanto, o custo computacional desses algoritmos limita a capacidade de resposta imediata às intrusões. Neste trabalho, é proposta uma arquitetura para a detecção e prevenção de intrusões em tempo real em switches eBPF a partir de modelos otimizados assincronamente através da meta-heurística GRASP-FS. Como prova de conceito, um modelo é construído a partir de um computador e atualizado periodicamente em um switch eBPF. Os resultados obtidos revelam que a solução proposta é capaz de detectar e prevenir intrusões em tempo real com baixa sobrecarga para os cenários avaliados.

Cardiac monitoring based on wearable photoplethysmography (PPG) is widespread because of its usability and low cost. Unfortunately, PPG is negatively affected by various types of disruptions, which could introduce errors to the algorithm that extracts pulse rate variability (PRV). This study aims to identify the nature of such artifacts caused by various types of factors under the conditions of precisely planned experiments. We also propose methods for their reduction based solely on the PPG signal while preserving the frequency content of PRV. The accuracy of PRV derived from PPG was compared to heart rate variability (HRV) derived from the accompanying ECG. The results indicate that filtering PPG signals using the discrete wavelet transform and its inverse (DWT/IDWT) is suitable for removing slow components and high-frequency noise. Moreover, the main benefit of amplitude demodulation is better preparation of the PPG to determine the duration of pulse cycles and reduce the impact of some other artifacts. Post-processing applied to HRV and PRV indicates that the correction of outliers based on local statistical measures of signals and the autoregressive (AR) model is only important when the PPG is of low quality and has no effect under good signal quality. The main conclusion is that the DWT/IDWT, followed by amplitude demodulation, enables the proper preparation of the PPG signal for the subsequent use of PRV extraction algorithms, particularly at rest. However, post-processing in the proposed form should be applied more in the situations of observed strong artifacts than in motionless laboratory experiments.

Este trabalho apresenta o aplicativo móvel MedBe, que conecta profissionais de saúde e bem estar (e.g., enfermagem, cuidador de idoso, psicologia, etc.) com pacientes que desejam agendar serviços. A principal inovação do MedBe é a possibilidade de agendamento de atendimentos sob demanda, através de um algoritmo de matching que aumenta a eficiência da busca para os pacientes, e amplia de maneira qualificativa o mercado para profissionais. Apresentamos as principais funcionalidades do aplicativo que está em operação e já conta com mais de 2.000 instalações e 350 solicitações. O público-alvo deste trabalho se beneficiará com insights de tecnologia e inovação no setor de saúde digital. Trabalhos futuros incluem a abordagem de aspectos de usabilidade, escalabilidade e implementação de técnicas de inteligência artificial.

The Emognition dataset is dedicated to testing methods for emotion recognition (ER) from physiological responses and facial expressions. We collected data from 43 participants who watched short film clips eliciting nine discrete emotions: amusement, awe, enthusiasm, liking, surprise, anger, disgust, fear, and sadness. Three wearables were used to record physiological data: EEG, BVP (2x), HR, EDA, SKT, ACC (3x), and GYRO (2x); in parallel with the upper-body videos. After each film clip, participants completed two types of self-reports: (1) related to nine discrete emotions and (2) three affective dimensions: valence, arousal, and motivation. The obtained data facilitates various ER approaches, e.g., multimodal ER, EEG- vs. cardiovascular-based ER, discrete to dimensional representation transitions. The technical validation indicated that watching film clips elicited the targeted emotions. It also supported signals’ high quality.

Wireless sensor networks (WSNs) are commonly adopted by the Industrial Internet-of-Things (IIoT) scenarios due to their easy and fast deployment. However, WSNs are especially vulnerable to security attacks as reflexive packet flooding Denial of Service (DoS) may occur. Admission control and reputation-based strategies are effective for attack detection. However, mitigation strategies to deal with the effects of these attacks, such as congestion channel transmission is an open issue. Additionally, the resource-constrained nature of nodes, such as the low bandwidth, limited memory, and battery within WSNs, poses a challenge to develop efficient mechanisms in such a scenario. To address this issue, we propose a distributed congestion control by duty-cycle restriction (D-ConCReCT) to detect and mitigate DoS in IIoT. The main goal is to investigate its feasibility in large-scale networks, as well as its ability to reduce both the detection and mitigation times when compared to a previous centralized approach, the so-called congestion control by duty-cycle restriction (ConCReCT). Our results indicate that D-ConCReCT may be applied in the mitigation of DoS attacks in a sensor network scenario with 500 nodes.

Najstarsze Bractwo w Wadowicach funkcjonowało przez prawie cały okres staropolski, od 1616 do 1822 roku. Skupiało 2231 wiernych z parafii wadowickiej Ofiarowania NMP oraz okolicznych miejscowości. Odtworzona historia ruchu brackiego na podstawie zaginionej dziś XVIII wiecznej Księgi Brackiej stanowi dużą wartość dla poznania religijności lokalnej społeczności. Informacje zawarte w pracy posłużyć mogą również religioznawcom, genealogom, teologom.

In our study, we examine the impact of citation network structures on the ability to discern valuable research topics in Computer Science literature. We use the bibliographic information available in the DBLP database to extract candidate phrases from scientific paper abstracts. Following that, we construct citation networks based on direct citation, co-citation and bibliographic coupling relationships between the papers. The candidate research topics, in the form of keyphrases and n-grammes, are subsequently ranked and filtered by a graph-text ranking algorithm. This selection of the highest ranked potential topics is further evaluated by domain experts and through the Wikipedia knowledge base. The results obtained from these citation networks are complementary, returning valid but non-overlapping output phrases between some pairs of networks. In particular, bibliographic coupling appears to capture more unique information than either direct citation or co-citation. These findings point towards the possible added value in combining bibliographic coupling analysis with other structures. At the same time, combining direct citation and co-citation is put into question. We expect our findings to be utilised in method design for research topic identification.

A graph of the changes in an engine’s operating speed can be used to assess the quality of the combustion in its cylinders. In this paper, the authors carried out tests on a Buckau-Wolf R8VD-136 ship engine, which was directly driving the propeller. This engine is owned by the Laboratory of Marine Engine Rooms at the Maritime University of Szczecin. For standard rotational speeds ranging from 200 to 280 rpm, with increments of 20 rpm, the authors measured the changes in the instantaneous speed for the engine’s normal operating conditions (reference graphs) as well as with one of the cylinders being out of operation. A no-combustion situation was successively introduced into each cylinder for each preset rotational speed. The obtained graphs of the instantaneous speed were then used to determine certain quantitative indicators, which the authors believe can provide information about the technical condition of the engine. The analysis concerned the averaged graphs of the speed under the conditions set for five consecutive engine operating cycles. The indicators that were calculated included the maximum difference in the speed over the engine’s full operating cycle, the uniformity of the engine speed and the differential speed area factor, the latter a term that has been proposed by the authors. The values of the individual indicators that were obtained from the reference graphs and the graphs with no combustion in one of the cylinders were compared. All indicators are sensitive to cylinder misfire. Conclusions were then drawn on the usefulness of these indicators in assessing the condition of an engine.

A high percentage of information that propagates through a social network is sourced from different exogenous sources. E.g., individuals may form their opinions about products based on their own experience or reading a product review, and then share that with their social network. This sharing then diffuses through the network, evolving as a combination of both network and external effects. Besides, different individuals (nodes in a social network) have different degrees of exposition to their external sources, as well. Modeling this influence of external sources is important in order to understand the diffusion process and predict future content sharing patterns. Recognizing this fusion of intrinsic (network) effect and exogenous (external) effect, this paper develops a novel fuzzy relative willingness (FRW) model. Leveraging a fuzzy set approach provides a way to handle the uncertainties arising within the human concept of willingness. We demonstrate that FRW is able to accurately identify both top- $k$ most content producers and diffusion effect based on external influence. We also demonstrate that the fuzzy set theory provides a compelling framework to model uncertainties pertaining to the influence as well as the susceptibility of individuals for both network and exogenous effects.

Electronic health record (EHR) systems play a relevant role in hospital daily activities. They support the access of patient data, such as medication delivery and history of hospitalizations. Currently, it is essential to provide easy and secure access to medical records by hospital staff. Pervasive and ubiquitous computing can lead to a better way to manage EHR systems. In this scenario, one of the greatest research challenges is to ensure device authentication avoiding its impersonation. The main contribution of this work is to carry out an extended experimental evaluation of a secure architecture based on ubiquitous computing for medical records retrieval and maintenance to demonstrate its feasibility through of experiments in laboratory and hospital scenarios. For evaluation, this article implements a prototype that performs the medication delivery procedure. Experimental results indicate the system functionality, usability with a decreasing user learning curve, and no significant overhead in terms of communication delay.

A Internet das Coisas é uma realidade tanto na indústria quanto na vida cotidiana das pessoas. Nesse contexto, informações sensíveis relacionadas a indivíduos, empresas e governos devem ser mantidas em sigilo. No entanto, a escolha de mecanismos criptográficos para essa finalidade deve considerar a disponibilidade de recursos dos dispositivos. Neste trabalho, apresentamos uma avaliação de diferentes mecanismos criptográficos simétricos para fornecer confidencialidade às mensagens dos protocolos MQTT e CoAP. São analisadas métricas como o consumo energético e o tempo de resposta. Experimentos práticos indicam que a escolha do algoritmo criptográfico pode levar a uma economia de até 32,29% de energia e uma redução de 41,60% no uso de CPU.

The introduction of the IPv6 protocol solved the problem of providingaddresses to network devices. With the emergence of the Internetof Things (IoT), there was also the need to develop a protocolthat would assist in connecting low-power devices. The 6LoWPANprotocols were created for this purpose. However, such protocolsinherited the vulnerabilities and threats related to Denial of Service(DoS) attacks from the IPv4 and IPv6 protocols. In this paper, weprepare a network environment for low-power IoT devices usingCOOJA simulator and Contiki operating system to analyze theenergy consumption of devices. Besides, we propose an IntrusionDetection System (IDS) associated with the AES symmetric encryptionalgorithm for the detection of reflection DoS attacks. Thesymmetric encryption has proven to be an appropriate methoddue to low implementation overhead, not incurring in large powerconsumption, and keeping a high level of system security. The maincontributions of this paper are: (i) implementation of a reflectionattack algorithm for IoT devices; (ii) implementation of an intrusiondetection system using AES encryption; (iii) comparison ofthe power consumption in three distinct scenarios: normal messageexchange, the occurrence of a reflection attack, and runningIDS algorithm. Finally, the results presented show that the IDSwith symmetric cryptography meets the security requirements andrespects the energy limits of low-power sensors.

O bjectives: Copeptin has been reported to play an important role in metabolic response in women with PCOS. However, the optimal cut-off value for detecting subjects with insulin resistance (IR) remains undetermined. We investigated whether copeptin can serve as an indicator of IR and tried to determine the optimal cut-off value of plasma copeptin concentration in detecting subjects with PCOS and IR. Material and methods: We carried out a case-control study on 158 women with PCOS and HOMA-IR < 2.5, 96 women with PCOS with HOMA-IR ≥ 2.5, and 70 healthy volunteers. Plasma copeptin, as well as hormonal, biochemical, metabolic, and IR parameters, were measured. To investigate whether copeptin allows IR to be predicted in PCOS, we used logistic regression models and ROC curve analysis. Results: Median plasma copeptin concentration was the highest in the women with PCOS and HOMA-IR ≥ 2.5. Logistic regression analysis revealed that copeptin was the strongest predictor of HOMA ≥ 2.5 (OR: 53.34 CI 7.94–358.23, p < 0.01). Analysis of ROC curves indicated that the cut-off value above 4 pmol/L of plasma copeptin concentration had high (99%) specificity but very low (21%) sensitivity in diagnosing of IR (AUC 0.607 (95% CI 0.53–0.68. Conclusions: Our findings suggest that copeptin is associated with IR in PCOS patients, but due to low sensitivity should not be considered as a marker of IR.

Allium scorodoprasum (Amaryllidaceae) is a vulnerable, red-listed species in Poland, where it has an uneven distribution. In the southwestern part of the country it has not been confirmed since 1961–1975 in the majority of the former areas of its occurrence. The paper describes 48 localities of A. scorodoprasum and provides data on population size, occupied habitats, plant communities and threats in the section of the Oder Valley between Oława, Wrocław and Brzeg Dolny in Lower Silesia (SW Poland).

Human communication is commonly represented as a temporal social network, and evaluated in terms of its uniqueness. We propose a set of new entropy-based measures for human communication dynamics represented within the temporal social network as event sequences. Using real world datasets and random interaction series of different types we find that real human contact events always significantly differ from random ones. This human distinctiveness increases over time and by means of the proposed entropy measures, we can observe sociological processes that take place within dynamic communities.

Emotion and affect recognition in the uncontrolled everyday-life environment remains challenging. One of its vital problems is collecting numerous annotated emotional samples indispensable for learning the reasoning model. We propose a novel method supporting rich emotional data collection - a pre-trained binary model recognizing physiologically arousing events in real-time and triggering self-assessments at a convenient point in time. An experimental study on 6.000 hours of recorded physiological signals has been performed. The results suggest that we are able to detect emotional events in real-life scenarios to enhance data collection for emotion recognition in the field.

Smartphones have become an integral part of our lives. One of their crucial functionalities is sharing data. We analyze the communication modules in Android devices (WiFi, Bluetooth, NFC) in terms of parallel data streaming capabilities. We find that increasing the number of concurrent threads reduces the broadcast time, but also consumes a lot of other resources, e.g. RAM. Furthermore, the linear increase in the number of threads created does not guarantee the linear decrease in transfer time. We also find that the Bluetooth 5.0 has, in fact, slower transmission rate than its predecessors.

There is content such as hate speech, offensive, toxic or aggressive documents, which are perceived differently by their consumers. They are commonly identified using classifiers solely based on textual content that generalize pre-agreed meanings of difficult problems. Such models provide the same results for each user, which leads to high misclassification rate observable especially for contentious, aggressive documents. Both document controversy and user nonconformity require new solutions. Therefore, we propose novel personalized approaches that respect individual beliefs expressed by either user conformity-based measures or various embeddings of their previous text annotations. We found that only a few annotations of most controversial documents are enough for all our personalization methods to significantly outperform classic, generalized solutions. The more controversial the content, the greater the gain. The personalized solutions may be used to efficiently filter unwanted aggressive content in the way adjusted to a given person.

Wearables equipped with pervasive sensors enable us to monitor physiological and behavioral signals in our everyday life. We propose the WellAff system able to recognize affective states for wellbeing support. It also includes health care scenarios, in particular patients with chronic kidney disease (CKD) suffering from bipolar disorders. For the need of a large-scale field study, we revised over 50 off-the-shelf devices in terms of usefulness for emotion, stress, meditation, sleep, and physical activity recognition and analysis. Their usability directly comes from the types of sensors they possess as well as the quality and availability of raw signals. We found there is no versatile device suitable for all purposes. Using Empatica E4 and Samsung Galaxy Watch, we have recorded physiological signals from 11 participants over many weeks. The gathered data enabled us to train a classifier that accurately recognizes strong affective states., Accepted to the International Workshop on Artificial Intelligence for Mobile and Ubiquitous Communication System, EAI MobiQuitous 2020

The Karkonosze National Park is affected by long-distance metal air transport of anthropogenic contamination as well as by tourist activity. Therefore, concentrations of Cd, Co, Cr, Cu, Fe, Mn, Ni, Pb and Zn were evaluated in soil as well as in vital and non-vital fronds of Athyrium distentifolium Opiz collected in the middle of the growing season from glacial cirques in the Karkonosze. Additionally, fronds of the same species turning brown in autumn were collected from the same sampling sites. The health of ferns was impacted by the contents of Co, Cu, Fe, Mn, Ni, Pb and Zn in fronds, which may indicate that A. distentifolium uses elements accumulated in its tissues as defence against pathogens. Individuals from higher altitudes contained higher concentrations of Cd, Cu, Ni, Fe and Pb than those from lower altitudes. Autumn fronds of A. distentifolium contained a significantly higher concentration of Cr, Cu, Fe, Mn and Pb than vital summer fronds, which may indicate sequestration of these elements in senescing fronds probably to remove potentially harmful metals. Non-vital ferns were a better accumulator of Pb than vital ones, while both types of ferns accumulated Mn in a similar way. Electronic supplementary material The online version of this article (10.1007/s10661-020-08438-4) contains supplementary material, which is available to authorized users.

Intrusion Detection Systems (IDS) are important tools to detect malicious network traffic. Particularly, the efficient detection of Denial of Service (DoS) attacks poses a challenging issue especially in Internet of Things (IoT) scenarios composed of resource-constrained devices, as Wireless Sensor Networks (WSN). Additionally, machine learning classification methods have been presented as a prominent approach to DoS detection. However, a lack of a suitable and careful assessment of such methods becomes difficult to understand the real contribution of them to improve the detection of DoS attacks in WSNs. This work aiming at assessing the efficiency of machine learning classification approaches to detect (i) flooding, (ii) gray hole, and (iii) black hole DoS attacks in WSNs. Our evaluation is based on a WSN-based dataset, called WSN-DS, considering the accuracy and speediness metrics. Results reveal that—considering accuracy and speediness simultaneously—J48 method is recommended for detecting gray hole and black hole attacks, whereas the Random Tree method is the best option for flooding detection. Regarding the speediness, the J48 method is the fastest consuming 0.54µs of processing on average per sample.

Wearables like smartwatches or wrist bands equipped with pervasive sensors enable us to monitor our physiological signals. In this study, we address the question whether they can help us to recognize our emotions in our everyday life for ubiquitous computing. Using the systematic literature review, we identified crucial research steps and discussed the main limitations and problems in the domain.

This article suggested broadening the standard methods for diagnosing the technical condition of diesel engines to include an analysis of the instantaneous rotational speed of compression combustion engines with the use of a novel algorithm. The authors revised the subject concerning the use of the analysis of instantaneous changes in the rotational speed of an engine when assessing its technical condition and the location of the malfunction. An algorithm and its practical implementation in a prototype diagnostic system called SpeedMA were presented. This article reported the test results of the prototype in the context of indicating the engine cylinder in which ignition failed to occur. Tests were carried out for two marine engines: a low-speed trunk engine directly driving the fixed-pitch propeller and a medium-speed trunk engine driving the alternator. For each case, an analysis was carried out for different engine loads and at individual cylinders in which combustion failed to occur. The experimental results showed an unambiguous relation between the combustion process of the examined engines and changes in the instantaneous rotational speed. The results also confirmed the usefulness of the proposed method and showed the correct operation of the presented diagnostic algorithm. The proposed diagnostic system could be used during the operation of engines running in real ship engine rooms.

Recently, a variety of model designs and methods have blossomed in the context of the sentiment analysis domain. However, there is still a lack of comprehensive studies of Aspect-based Sentiment Analysis. We want to fill this gap and propose a comparison with ablation analysis of Aspect Term Extraction using various text embeddings methods. We particularly focused on simple architectures based on long short-term memory (LSTM) with optional conditional random field (CRF) enhancement using different pre-trained word embeddings. Moreover, we analyzed the influence on the performance of extending the word vectorization step with character-based word embeddings. The experimental results on SemEval datasets revealed that bi-directional long short-term memory (BiLSTM) could be used as a very good predictor, even comparing to very sophisticated and complex models using huge word embeddings or language models. We presented a comprehensive analysis of various customizations of LSTM-based architecture and word/character embeddings that could be used as a guideline to choose the best model version for particular user needs.

Analysis of subjective texts like offensive content or hate speech is a great challenge, especially regarding annotation process. Most of current annotation procedures are aimed at achieving a high level of agreement in order to generate a high quality reference source. However, the annotation guidelines for subjective content may restrict the annotators’ freedom of decision making. Motivated by a moderate annotation agreement in offensive content datasets, we hypothesize that personalized approaches to offensive content identification should be in place. Thus, we propose two novel perspectives of perception: group-based and individual. Using demographics of annotators as well as embeddings of their previous decisions (annotated texts), we are able to train multimodal models (including transformer-based) adjusted to personal or community profiles. Based on the agreement of individuals and groups, we experimentally showed that annotator group agreeability strongly correlates with offensive content recognition quality. The proposed personalized approaches enabled us to create models adaptable to personal user beliefs rather than to agreed offensiveness understanding. Overall, our individualized approaches to offensive content classification outperform classic data-centric methods that generalize offensiveness perception and it refers to all six tested models. Additionally, we developed requirements for annotation procedures, personalization and content processing to make the solutions human-centered.

Scientific output, as measured in research published annually, has seen a consistent growth for decades now. As more manuscripts are submitted for publication each year, new publishing venues appear – often as increasingly specialised offshoots of existing journals and conferences. This situation presents scholars with a wealth of publishing venues to consider and choose from for their manuscripts. Surprisingly, we find that the most cited papers are not necessarily found in the highest-ranked journals, while the best conferences dominate in this regard. We find it intriguing that popular Computer Science conferences act like a vacuum of attention, centralising all good publications, while journals are carried less by their renown and thus can attract strong manuscripts even at a low rank. But to what extent does a venue imply a paper’s recognition and popularity? We propose a new approach and processing pipeline, in which we project a heterogeneous publication network, learn representations for papers and classify said papers according to their 2-year citation success. This framework provides a groundwork for a venue recommendation system, a tool for improving paper’s scientific recognition through a better choice of a publishing venue.