Search

Newly emerging and evolving technologies such as Cloud, Fog and Edge Computing, as well as Internet of Things, Cyber-Physical Systems and Distributed Ledger Technology (such as blockchain) together with advances in Artificial Intelligence (AI) research are increasingly becoming a common and pervasive phenomenon in our everyday lives. Their co-evolution with society is driving the emergence of future socio-technical systems, which further promote ubiquitous entanglement between humans and machines. Fog, Edge and Dew computing as post-Cloud computing paradigms aim to relocate computing resources closer to end users in order to mitigate cloud-specific issues of highly centralized computation. Dew computing as the youngest of the post-cloud paradigms promotes human centered independence and collaboration between devices within scalable distributed computing infrastructures. Meanwhile, the field of artificial intelligence is adapting to recent challenges posed by user data privacy regulations as well as opportunities for applications on mobile devices based on their growing computational abilities. The usage of artificial intelligence in pervasive and scalable distributed computing systems is a natural step towards ubiquitous intelligent infrastructures and collaborative human and machine environments. Federated learning is an artificial intelligence technique enabling collaborative learning in distributed devices environment without sharing the training data sets, which are often private. This paper provides the overview of the federated learning paradigm showing that it inherently leverages both independence and collaboration, thus exemplifying implementation of dew intelligence within scalable distributed computing hierarchy.

Spike glycoprotein is essential for the reproduction of the SARS-CoV-2 virus, and its inhibition using already approved antiviral drugs may open new avenues for the treatment of patients with the COVID-19 disease. Because of that, we analyzed the inhibition of SARS-CoV-2 spike glycoprotein with FDA-approved antiviral drugs and their double and triple combinations. We used the Vini in silico model of cancer to perform this virtual drug screening, showing HIV drugs to be the most effective. Besides, the combination of cobicistat-abacavir-rilpivirine HIV drugs demonstrated the highest in silico efficacy of inhibiting SARS-CoV-2 spike glycoprotein. Therefore, a clinical trial of cobicistat- abacavir-rilpivirine on a limited number of COVID- 19 patients in moderately severe and severe conditions is warranted.

Available online 30 April 2021. Brainhack is an innovative meeting format that promotes scientific collaboration and education in an open, inclusive environment. This NeuroView describes the myriad benefits for participants and the research community and how Brainhacks complement conventional formats to augment scientific progress. The present manuscript is part of a growing community effort to collate Brainhack-related insights and expertise into a Jupyter Book (http://brainhack.org/brainhack_jupyter_book/) that will serve as a centralized set of resources for the community; we acknowledge all the individuals who contributed and will make ongoing contributions to these resources. A pre-print version of the present manuscript is available as part of the Jupyter Book. Moreover, we would like to acknowledge all Brainhack organizers, supporters, presenters, and participants for their contribution to growing and maintaining this community. The benefits described in this manuscript would not be possible without them. We also thank all institutions, labs, and organizations who have helped this community grow, meet in stimulating environments, and add an excellent educational resource pool and agenda. With an expanding community, Brainhack’s support network keeps growing, and we thank all labs and individual researchers for their dedication and expertise offered to this community (see http://brainhack.org/brainhack_jupyter_book/acknowledgments.html for a full list of individual acknowledgments; an updated list will be maintained in the Jupyter Book). Grants and funding bodies: This project has received funding from the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement no. 867458 awarded to Julia M. Huntenburg; from ANR-19-DATA-0025 NeuroWebLab for Katja Heuer, Roberto Toro, and Nicholas Traut; and from NIH K00MH122372 for Stephanie Noble. The Brainhack Community member list and contributions of the different authors are detailed at http://brainhack.org/brainhack_jupyter_book/contributors.html. Our crediting system is described here: http://brainhack.org/brainhack_jupyter_book/neuroview_authorship-agreement.html.

Virtual drug screening is one of the most widely used approaches for finding new drugs candidates. The process consists in selecting one or more chemical compounds with the highest binding free energy to target proteins. Given that the empirical space of chemical compounds is extremely large and estimated to has over 50 millions of them, finding the most effective drug is computationally challenging. Furthermore, the vast majority of proteins still lack the experimentally obtained 3D structures, making it hard to accurately calculate their binding free energies with chemical compounds. With this in mind, the aim of our study was to investigate the accuracy of the Autodock Vina program in a virtual drug screening on a set of proteins that do not have experimentally determined structures. To do this, we performed a virtual drug screening with the Autodock Vina on a large set of drug-kinase pairs taken from the IDG-Dream Drug-Kinase Binding Prediction Challenge. The results obtained show that the Autodock Vina can be used effectively in such unstructured environments.

Machines usually employ a guess-and-check strategy to analyze data: they take the data, make a guess, check the answer, adjust it with regard to the correct one if necessary, and try again on a new data set. An active learning environment guarantees better performance while training on less, but carefully chosen, data which reduces the costs of both annotating and analyzing large data sets. This issue becomes even more critical for deep learning applications. Human-like active learning integrates a variety of strategies and instructional models chosen by a teacher to contribute to learners&rsquo, knowledge, while machine active learning strategies lack versatile tools for shifting the focus of instruction away from knowledge transmission to learners&rsquo, knowledge construction. We approach this gap by considering an active learning environment in an educational setting. We propose a new strategy that measures the information capacity of data using the information function from the four-parameter logistic item response theory (4PL IRT). We compared the proposed strategy with the most common active learning strategies&mdash, Least Confidence and Entropy Sampling. The results of computational experiments showed that the Information Capacity strategy shares similar behavior but provides a more flexible framework for building transparent knowledge models in deep learning.

Understanding the historical transformation of artistic styles implies the recognition of different stylistic properties. From a computer vision perspective, stylistic properties represent complex image features. In our work we explore the use of convolutional neural networks for learning features that are relevant for understanding properties of artistic styles. We focus on stylistic properties described by Heinrich Wolfflin in his book Principles of Art History (1915). Wolfflin identified five key visual principles, each defined by two contrasting concepts. We refer to each principle as one high-level image feature that measures how much each of the contrasting concepts is present in an image. We introduce convolutional neural network regression models trained to predict values of the five Wolfflin’s features. We provide quantitative and qualitative evaluations of those predictions, as well as analyze how the predicted values relate to different styles and artists. The outcome of our analysis suggests that the models learn to discriminate meaningful features that correspond to the visual characteristics of concepts described by Wolfflin. This indicates that the presented approach can be used to enable new ways of exploring fine art collections based on image features relevant and well-known within art history.

Recent research in machine learning pointed to the core problem of state-of-the-art models which impedes their widespread adoption in different domains. The models' inability to differentiate between noise and subtle, yet significant variation in data leads to their vulnerability to adversarial perturbations that cause wrong predictions with high confidence. The study is aimed at identifying whether the algorithms inspired by biological evolution may achieve better results in cases where brittle robustness properties are highly sensitive to the slight noise. To answer this question, we introduce the new robust gradient descent inspired by the stability and adaptability of biological systems to unknown and changing environments. The proposed optimization technique involves an open-ended adaptation process with regard to two hyperparameters inherited from the generalized Verhulst population growth equation. The hyperparameters increase robustness to adversarial noise by penalizing the degree to which hardly visible changes in gradients impact prediction. The empirical evidence on synthetic and experimental datasets confirmed the viability of the bio-inspired gradient descent and suggested promising directions for future research. The code used for computational experiments is provided in a repository at https://github.com/yukinoi/bio_gradient_descent.

The adaptation of deep learning models within safety-critical systems cannot rely only on good prediction performance but needs to provide interpretable and robust explanations for their decisions. When modeling complex sequences, attention mechanisms are regarded as the established approach to support deep neural networks with intrinsic interpretability. This paper focuses on the emerging trend of specifically designing diagnostic datasets for understanding the inner workings of attention mechanism based deep learning models for multivariate forecasting tasks. We design a novel benchmark of synthetically designed datasets with the transparent underlying generating process of multiple time series interactions with increasing complexity. The benchmark enables empirical evaluation of the performance of attention based deep neural networks in three different aspects: (i) prediction performance score, (ii) interpretability correctness, (iii) sensitivity analysis. Our analysis shows that although most models have satisfying and stable prediction performance results, they often fail to give correct interpretability. The only model with both a satisfying performance score and correct interpretability is IMV-LSTM, capturing both autocorrelations and crosscorrelations between multiple time series. Interestingly, while evaluating IMV-LSTM on simulated data from statistical and mechanistic models, the correctness of interpretability increases with more complex datasets.

Fundamental research in physics has long been a prerequisite for computer scientists and engineers to design innovative products, such as laptops and cell phones. Technological innovations and fundamental research are both part of the so-called STEM fields (Science, Technology, Engineering, and Mathematics), which are known to substantially contribute to economic growth. However, the questions still remain: how much contribution do these fields make to both wealth accumulation and inequality at different levels of analysis? First, analyzing the lists of world’s wealthiest individuals, the Zipf plot analysis demonstrates that STEM billionaires contribute more to wealth inequality than their non-STEM counterparts. Analyzing the companies in the S&P500, we find that STEM firms contribute more to wealth inequality and have larger growth rates on average than the non-STEM firms. Finally, we show that the more STEM graduates in a country, the larger its GDP growth rate. In combination, we demonstrate that STEM is a fractal mechanism that drives wealth accumulation—and the wealth inequality— at different scales of economy—from individual wealth to firm valuation to country GDP.

The increasing availability of large digitized fine art collections opens new research perspectives in the intersection of artificial intelligence and art history. Motivated by the successful performance of Convolutional Neural Networks (CNN) for a wide variety of computer vision tasks, in this paper we explore their applicability for art-related image classification tasks. We perform extensive CNN fine-tuning experiments and consolidate in one place the results for five different art-related classification tasks on three large fine art datasets. Along with addressing the previously explored tasks of artist, genre, style and time period classification, we introduce a novel task of classifying artworks based on their association with a specific national artistic context. We present state-of-the-art classification results of the addressed tasks, signifying the impact of our method on computational analysis of art, as well as other image classification related research areas. Furthermore, in order to question transferability of deep representations across various source and target domains, we systematically compare the effects of domain-specific weight initialization by evaluating networks pre-trained for different tasks, varying from object and scene recognition to sentiment and memorability labelling. We show that fine-tuning networks pre-trained for scene recognition and sentiment prediction yields better results than fine-tuning networks pre-trained for object recognition. This novel outcome of our work suggests that the semantic correlation between different domains could be inherent in the CNN weights. Additionally, we address the practical applicability of our results by analysing different aspects of image similarity. We show that features derived from fine-tuned networks can be employed to retrieve images similar in either style or content, which can be used to enhance capabilities of search systems in different online art collections.

Faster and cheaper – those two words have driven the computer industry for five decades, and the rest of the economy has been along for the ride. Computer scientists and their collaborators have been forced to discard certain mental barriers and open their minds to radical change. A network is a generic topology of the basis of biological evolutionary development and technical functional processes. With the evolution of the motion of matter the process of shaping the network structure was initiated, because only thus is it possible to maintain and develop any distributed intelligent system. Generic certification of that obtained from the evolution of biological systems which based on set of networks (bloodstream networks, neural networks ....). The theory of signals, communications and computing was founded on the special motion of matter (charge carriers-electrons) extensively througth the networks. Therefore, communication dynamic compatibilization of networks speed with processes speed (G-phenomena) is the basis for distributed systems. This paper presents and defines the G-phenomena as a Millennium cornerstone of the rapid development of distributed systems on a scalability platform within the Grid and Cloud computing paradigm, offering a distributed platform management architecture designed for scalability, availability, usability, and high performance.

Real-world attacks can be interpreted as the result of competitive interactions between networks, ranging from predator-prey networks to networks of countries under economic sanctions. Although the purpose of an attack is to damage a target network, it also curtails the ability of the attacker, which must choose the duration and magnitude of an attack to avoid negative impacts on its own functioning. Nevertheless, despite the large number of studies on interconnected networks, the consequences of initiating an attack have never been studied. Here, we address this issue by introducing a model of network competition where a resilient network is willing to partially weaken its own resilience in order to more severely damage a less resilient competitor. The attacking network can take over the competitor nodes after their long inactivity. However, due to a feedback mechanism the takeovers weaken the resilience of the attacking network. We define a conservation law that relates the feedback mechanism to the resilience dynamics for two competing networks. Within this formalism, we determine the cost and optimal duration of an attack, allowing a network to evaluate the risk of initiating hostilities., 8 two-column pages, 6 figures, supplementary material; accepted for publication in Journal of the Royal Society Interface

Estimating the critical points at which complex systems abruptly flip from one state to another is one of the remaining challenges in network science. Due to lack of knowledge about the underlying stochastic processes controlling critical transitions, it is widely considered difficult to determine the location of critical points for real-world networks, and it is even more difficult to predict the time at which these potentially catastrophic failures occur. We analyse a class of decaying dynamic networks experiencing persistent failures in which the magnitude of the overall failure is quantified by the probability that a potentially permanent internal failure will occur. When the fraction of active neighbours is reduced to a critical threshold, cascading failures can trigger a total network failure. For this class of network we find that the time to network failure, which is equivalent to network lifetime, is inversely dependent upon the magnitude of the failure and logarithmically dependent on the threshold. We analyse how permanent failures affect network robustness using network lifetime as a measure. These findings provide new methodological insight into system dynamics and, in particular, of the dynamic processes of networks. We illustrate the network model by selected examples from biology, and social science.

The increasing energy consumption of large-scale high performance resources raises technical and economical concerns. A reduction of consumed energy in multicore systems is possible to some extent with an optimized usage of computing and memory resources that is tailored to specific HPC applications. The essential step towards more sustainable consumption of energy is its reliable measurements for each component of the system and selection of optimally configured resources for specific applications. This paper briefly surveys the current approaches for measuring and profiling power consumption in large scale systems. Then, a practical case study of a real-time power measurement of multicore computing system is presented on two real HPC applications: maximum clique algorithm and numerical weather prediction model. We assume that the computing resources are allocated in a HPC cloud on a pay-per-use basis. The measurements demonstrate that the minimal energy is consumed when all available cores (up to the scaling limit for a particular application) are used on their maximal frequencies and with threads binded to the cores.

With recent technological developments, modern computing systems are becoming more and more powerful. Common computers are capable of calculating more than 50, 000 million instructions per second. Currently the biggest obstacle in computer usage is slow communication between the human and the computer. With innovation and development of novel (natural based) sensors capable of capturing positions and natural motion of human body a new, natural, way of interacting with computer is enabled. This paper gives an overview of eye tracking technology, and presents its application in nature based human- computer interaction (HCI) systems by extending standard controllers, such as the keyboard and the mouse. Novel approaches enable HCI closer to the communication patterns of human beings. In addition, the usage of those concepts in applications for eye diseases and disorders diagnostics and rehabilitation is discussed.

The ever-increasing data production and availability in the field of bioinformatics demands a paradigm shift towards the utilization of novel solutions for efficient data storage and processing, such as the MapReduce data parallel programming model and the corresponding Apache Hadoop framework. Despite the evident potential of this model and existence of already available algorithms and applications, especially for batch processing of large data sets as in the Next Generation Sequencing analysis, bioinformatics MapReduce applications are yet to become widely adopted in the bioinformatics data analysis. We identify two prerequisites for their adaptation and utilization: (1) the ability to compose complex workflows from multiple bioinformatics MapReduce tools that will abstract technical details of how those tools are combined and executed allowing bioinformatics domain experts to focus on the analysis, and (2) the availability of accessible and flexible computing infrastructure for this type of data processing. This paper presents integration of two existing systems: Cloudgene, a bioinformatics MapReduce workflow framework, and CloudMan, a cloud manager for delivering application execution environments. Together, they enable delivery of bioinformatics MapReduce applications in the Cloud.

In this paper we investigate scalability issues of self-synchronization emergent properties, described with the pulse coupled oscillator model. As in the pulse coupled oscillator model the information propagation process is a gossip-like process, huge amounts of network traffic can be generated, causing thus scalability issues of the whole collective adaptive systems. These issues are even more emphasized in collective adaptive heterogeneous systems called Machine-to-Machine (M2M) systems. Namely, these systems consist not only from one large complex network, but also from a larger number of different interconnected complex networks. The easiest way to reduce network traffic in large networks is to use different overlay network topologies. An overlay network topology can be seen as a layer of a virtual network topology on top of a physical network, enabling significantly less messages to be exchanged during a synchronization process. However, the implementation process of overlay network topologies is not very efficient in real- world environments, as will be discussed in the paper. Therefore, we propose a mechanism for selective coupling implemented on the sender side that can be used to reduce both network traffic and time to synchronization without negatively affecting the entire synchronization process. Moreover, in some cases the rate of successful synchronization outcomes can be also increased when using the proposed mechanism.

Marine traffic and rapidly growing tourism in the Adriatic Sea region are the main reasons for investigating different weather phenomena and developing the prediction models that depend on distributed computing infrastructures. Weather phenomena like storms and waterspouts are common in the in the Mediterranean Basin and thus in the Adriatic Sea region. Their occurrence is becoming easier to follow in past few years thanks to the advanced numerical weather prediction models and modern communication and computing technologies. This paper presents AdriaScience gateway, the application specific gateway for meteorological applications based on WS-PGRADE/gUSE technology developed within the SCI-BUS project. The main goal of the application specific gateway is to facilitate a universal access to the different Croatian distributed computing infrastructures for Croatian meteorology community.

The ever-increasing energy consumption in large- scale distributed computing systems such as clusters, grids and clouds raises social, technical, economical, and environmental concerns. Therefore, designing novel energy-efficient approaches, to reduce energy consumption, at all levels of distributed system architecture is of great importance for the whole society. However, the essential step towards the introduction of energy-efficiency in largescale distributed systems is to measure the power consumption, accurately, reliably, and continually in each component of the system. This paper briefly surveys the current approaches for measuring and profiling power consumption in large scale distributed systems. Furthermore, the practical case study of a real- time power measurement in multi-core computing system, as a basic building block of a distributed computing system, is presented.

In this paper, a novel multi-resolution real-time 3D thermal imaging system as potential solution for a human body 3D thermal models standardisation is presented. The system consists of a high-resolution offline 3D scanner and a real-time low-resolution 3D scanner, both of them paired together with a thermal imaging camera. The emphasis of this paper is the presentation of the novel concept of the standardisation of human body 3D thermal models, captured by the multi-resolution real-time 3D thermal imaging system. The standardisation procedure utilises skeleton detection, skeleton transformation, mesh optimisation and texture mapping. The presented concept enables novel and practical methods for human body 3D thermal models comparison and analysis.

At the end of 2010, we are at the effective end of the second phase of research in the field of Social Networks (SNs) and aspects such as Human-to-Human (H2H) interactions have pretty much had their day due to advances in Machine-to-Machine (M2M) interactions. This chapter will provide a useful insight into the differences between those two types of SNs: the human SNs (hSNs) based on H2H interactions and the machine SNs (mSNs) based on M2M interactions. During the last two decades rapid improvements in computing and communication technologies have enabled a proliferation of hSNs and we believe they will induce the formation of mSNs in the next decades. To this end, we will show how to carry out successful SN analyses (e.g. clustering and data diffusion) by connecting ethological approaches to social behaviour in animals (e.g. the study of firefly synchronization) and M2M interactions.

All objects emit infrared radiation at differing levels, depending on their temperature. Infrared thermal imaging is the technique of producing an image from infrared radiation. This thermal image represents two-dimensional distribution of the infrared radiation emitted by object displaying the object's temperatures. High performance computing data centres deploy high-density blade servers that have high power and cooling requirements. Thermal management techniques are needed in different levels to ensure data centre's reliability and economical operating cost. Appropriately designed and accurate temperature monitoring is essential for every thermal management techniques. Infrared thermal imaging can provide a fine-resolution thermal image of blade systems deployed in data canter. This paper outlines an application of infrared thermal imaging in blade based data centres. For this purpose thermal monitoring of computer cluster at the blade-level is conducted using infrared thermal imaging monitoring system at the blade enclosure-level.

The dissipation of thermal radiation can be observed using thermal infrared cameras which generate images based on the amount of input radiation belonging to a small part of the electromagnetic spectrum (with wavelengths from 7 μmto 15 μm). Since thermal imaging is a simple, contactless, non-invasive and inexpensive imaging method, it is widely applicable in industry, medicine and research. The most common type of thermal imaging involves taking and analyzing only a single thermal image, and it is thus called static thermal imaging. In cases when a thermal process cannot be approximated as static, dynamic thermal imaging and analysis are applied. The idea of combining thermal imaging with 3D scanning methods has spawned in the last few years. 3D thermal imaging can have many applications and purposes, ranging from thermogram rectification to the creation of standardized 3D thermal models of various subjects that can later be used for comparison and evaluation. Although 3D thermal imaging systems exist, all of the examined ones were targeted on the acquisition and analysis of static 3D thermal models. This paper presents the development of a 4D thermography system through integration of dynamic 3D scanning and thermographic imaging, additionally providing markerless motion analysis, which together enable practical, non-invasive, accurate and automatic monitoring of the temperature changes in the human body, and the characterization of human motion. The workflow of the designed concept is outlined, and the components of the constructed system are thoroughly explained. The process of calibration of the system is described, as well as the methods of motion detection and analysis. Great emphasis is given on the possible medical applications of a 4D thermography system, such as medical diagnostics, human locomotive system rehabilitation and health status monitoring over prolonged time periods.

Infrared (IR) thermography determines the surface temperature of an object or human body using thermal IR measurement camera. It is an imaging technology which is contactless and completely non-invasive. These propertiesmake IR thermography a useful method of analysis that is used in various industrial applications to detect, monitor and predict irregularities in many fields from engineering to medical and biological observations. This paper presents a conceptual model of Medical 3D Thermography which introduces standardised 3D thermogram creation, representation and analysis concepts useful for variety of medical applications. The creation of 3D thermograms is possible through combining 3D scanning methods with thermal imaging.We describe development of a 3D thermography system integrating passive thermal imaging with 3D geometrical data from active 3D scanner.We outline the potential benefits of this system in medical applications. In particular, we emphasize the benefits of using this system for preventive detection of breast cancer.

In an Artificial Neural Network (ANN) a large number of highly interconnected simple nonlinear processing units work in parallel to solve a specific problem. Parallelism, modularity and dynamic adaptation are three characteristics typically associated with ANNs. Field Programmable Gate Array (FPGA) based reconfigurable computing architectures are well suited to implement ANNs as one can exploit concurrency and rapidly reconfigure to adapt the weights and topologies of an ANN. ANNs are suitable for and widely used in various real-life applications. A large portion of these applications are realized as embedded computer systems. With continuous advancements in VLSI technology FPGAs have become more powerful and power efficient, enabling the FPGA implementation of ANNs in embedded systems. This paper proposes an FPGA ANN framework which facilitates implementation in embedded systems. A case study of an ANN implementation in an embedded fall detection system is presented to demonstrate the advantages of the proposed framework.

Creating modular behaviours in JADE using a traditional method such as the FSM (Finite State Machine) can be a difficult task to achieve. The first issue with FSMs is that they do not allow reusability of logic in different contexts. Secondly, the FSMs do not lend themselves well to concurrency within the execution thread and thus eliminating the possibility for parallel behaviours. Lastly, as the number of states in FSMs becomes increasing large, it becomes difficult to manage them. In this paper we introduce our JBehaviourTrees Framework that extends JADE Behaviours with BTs (Behaviour Trees) model. BTs are built via the composition of basic tasks increasing the possibility for modularity and code reuse. The proposed approach is verified through a case study concerning a FIPA-Request Interaction Protocol.

Modern advances in technology allow new telemonitoring systems for prevention, early diagnosis and management of chronic and degenerative conditions. These remote monitoring systems reduce the need for recurring visits to hospital, allow physicians better insight into the patient’s state and help determine necessary level of care for each individual patient. Telemonitoring can also be applied on a long-term basis to elderly persons to detect gradual deterioration in their health status, which may imply a reduction in their ability to live independently. Functional potential of the upper extremities is a good indicator of how well a person can live independently and also helps determine their health status. This article presents OSGi residential gateway based telemonitring system. In addition, two new electronic devices used for monitoring functional degradation of upper extremities are introduced into the presented system. One of the two devices records dynamic properties of the hand grip (e.g. grip speed, grip stability, grip endurance) and the other device records hand movement dynamic properties (e.g. seed of movement, spectral components).

Cloud computing has revolutionized how availability and access to computing and storage resources is realized ; it has made it possible to provision a large computational infrastructure in a matter of minutes, all through a web browser. What it has not yet solved is accessibility of a tool execution environment where tools and data can easily be added and used in non-trivial scenarios. In this paper, we demonstrate how CloudMan (http://usecloudman.org) can be used to provide complete and complex tool execution environments for making cloud resources functional for a desired domain.