Search

Technology is shaping our lives in a multitude of ways. This is fuelled by a technology infrastructure, both legacy and state of the art, composed of a heterogeneous group of hardware, software, services, and organisations. Such infrastructure faces a diverse range of challenges to its operations that include security, privacy, resilience, and quality of services. Among these, cybersecurity and privacy are taking the centre-stage, especially since the General Data Protection Regulation (GDPR) came into effect. Traditional security and privacy techniques are overstretched and adversarial actors have evolved to design exploitation techniques that circumvent protection. With the ever-increasing complexity of technology infrastructure, security and privacy-preservation specialists have started to look for adaptable and flexible protection methods that can evolve (potentially autonomously) as the adversarial actor changes its techniques. For this, Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL) were put forward as saviours. In this paper, we look at the promises of AI, ML, and DL stated in academic and industrial literature and evaluate how realistic they are. We also put forward potential challenges a DL based security and privacy protection system has to overcome. Finally, we conclude the paper with a discussion on what steps the DL and the security and privacy-preservation community have to take to ensure that DL is not just going to be hype, but an opportunity to build a secure, reliable, and trusted technology infrastructure on which we can rely on for so much in our lives.

Patent terminology relation extraction is of great significance to the construction of patent Knowledge graph. In order to solve the problem of long-distance dependency in traditional depth learning, a new method of patent terminology relation extraction is proposed, which combines attention mechanism and bi-directional LSTM model and with keyword strategy. Category keyword features in each sentence obtained by the improved TextRank with the patent text information vectorization added. BiLSTM neural work and attention mechanism are employed to extract the temporal information and sentence-level global feature information. Moreover, pooling layer is added to obtain the local features of the text. Finally, we fuse the global features and local features, and output the final classification results through the softmax classifier. The addition of category keywords improves the distinction of categories. Substantial experimental results demonstrate that the proposed model outperform the state-of-art neural model in patent terminology relation extraction.

This paper shows the automatic classification of research papers published in Scopus. Our classification is based on the research lines of the university. We apply the K-nearest neighbor classifier and linear discriminant analysis (LDA). Various stages were used from information gathering, creating the vocabulary, pre-processing and data training, and supervised classification in this work. The experiment involved 596 research articles published in SCOPUS from 2003–2017. The results show an overall accuracy of 88.44%.

Coreference resolution is essential for automatic text understanding to facilitate high-level information retrieval tasks such as text summarisation or question answering. Previous work indicates that the performance of state-of-the-art approaches (e.g. based on BERT) noticeably declines when applied to scientific papers. In this paper, we investigate the task of coreference resolution in research papers and subsequent knowledge graph population. We present the following contributions: (1) We annotate a corpus for coreference resolution that comprises 10 different scientific disciplines from Science, Technology, and Medicine (STM); (2) We propose transfer learning for automatic coreference resolution in research papers; (3) We analyse the impact of coreference resolution on knowledge graph (KG) population; (4) We release a research KG that is automatically populated from 55,485 papers in 10 STM domains. Comprehensive experiments show the usefulness of the proposed approach. Our transfer learning approach considerably outperforms state-of-the-art baselines on our corpus with an F1 score of 61.4 (+11.0), while the evaluation against a gold standard KG shows that coreference resolution improves the quality of the populated KG significantly with an F1 score of 63.5 (+21.8).

Machine learning is widely used in malware detection systems as a core component. However, machine learning algorithm is based on the assumption that the underlying malware concept is stable for training and testing. The assumption is vulnerable to well-crafted concept drift attacks, such as mimicry attacks, gradient descent attacks, poisoning attacks and so on. This paper proposes an ensemble learning system which combines vertical and horizontal correlation learning models. The significant diversity among vertical and horizontal correlation models increases the difficulty of concept drift attacks. And average p-value assessment is applied to fortify the system to be sensitive to hidden concept drift. The experiment results show that the hybrid system could actively recognize the concept drift among different Miuref variants.

Co-citation analysis can be exploited as a bibliometric technique used for mining information on the relationships between scientific papers. Proposed methods rely, however, on co-citation counting techniques that slightly take the semantic aspect into consideration. The present study proposes a new technique based on the measure of Semantic Similarity (SS) between the titles of co-cited papers. Several computational measures rely on knowledge resources to quantify the semantic similarity, such as the WordNet «is a» taxonomy. Our proposal analyzes the SS between the titles of co-cited papers using word-based SS measures. Two major analytical experiments are performed: the first includes the benchmarks designed for testing word-based SS measures; the second exploits the dataset DBLP (DBLP: Digital Bibliography & Library Project.) citation network. As a result, we found the SS measures behave the same as human judgement for the lexical similarity and can be consequently used for the automatic assessment of similarity between co-cited papers. The analysis of highly repeated co-citations demonstrates that the different SS measures display almost similar behaviours, with slight differences due to the distribution of the provided SS values. Furthermore, we note a low percentage of similar referred papers into the co-citations.

Understanding the current research trends, problems, and their innovative solutions remains a bottleneck due to the ever-increasing volume of scientific articles. In this paper, we propose NLPExplorer, a completely automatic portal for indexing, searching, and visualizing Natural Language Processing (NLP) research volume. NLPExplorer presents interesting insights from papers, authors, venues, and topics. In contrast to previous topic modelling based approaches, we manually curate five course-grained non-exclusive topical categories namely Linguistic Target (Syntax, Discourse, etc.), Tasks (Tagging, Summarization, etc.), Approaches (unsupervised, supervised, etc.), Languages (English, Chinese, etc.) and Dataset types (news, clinical notes, etc.). Some of the novel features include a list of young popular authors, popular URLs and datasets, list of topically diverse papers and recent popular papers. Also, it provides temporal statistics such as yearwise popularity of topics, datasets, and seminal papers. To facilitate future research and system development, we make all the processed dataset accessible through API calls. The current system is available at http://nlpexplorer.org.

This paper is a plea for developing possibilistic learning methods that would be consistent with if-then rule-based reasoning. The paper first recall the possibility theory-based handling of cascading sets of parallel if-then rules. This is illustrated by an example describing a classification problem. It is shown that the approach is both close to a possibilistic logic handling of the problem and can also be put under the form of a max-min-based matrix calculus describing a function underlying a structure somewhat similar to a max-min neural network. The second part of the paper discusses how possibility distributions can be obtained from precise or imprecise statistical data, and then surveys the few existing works on learning in a possibilistic setting. A final discussion emphasizes the interest of handling learning and reasoning in a consistent way.

Paper manufacturing is energy demanding and improved modelling of the pulp bleach process is the main non-invasive means of reducing energy costs. In this paper, time it takes to bleach paper pulp to desired brightness is examined. The model currently used is analysed and benchmarked against two machine learning models Random Forest and TreeBoost. Results suggests that the current model can be superseded by the machine learning models and it does not use the optimal compact subset of features. Despite the differences between the machine learning models, a feature ranking correlation has been observed for the new models. One novel, yet unused, feature that both machine learning models found to be important is the concentration of bleach agent.

Scientific papers are important for scholars to track trends in specific research areas. With the increase in the number of scientific papers, it is difficult for scholars to read all the papers to extract emerging or noteworthy knowledge. Paper modeling can help scholars master the key information in scientific papers, and relation classification (RC) between entity pairs is a major approach to paper modeling. To the best of our knowledge, most of the state-of-the-art RC methods are using entire sentence’s context information as input. However, long sentences have too much noise information, which is useless for classification. In this paper, a flexible context is selected as the input information for convolution neural network (CNN), which greatly reduces the noise. Moreover, we find that entity type is another important feature for RC. Based on these findings, we construct a typical CNN architecture to learn features from raw texts automatically, and use a softmax function to classify the entity pairs. Our experiment on SemEval-2018 task 7 dataset yields a macro-F1 value of 83.91%, ranking first among all participants.

E-commerce customers and general Web users should not believe that the products suggested by Recommender systems or results displayed by Web search engines are either complete or relevant to their search aspirations. The economic priority of Web related businesses requires a higher rank on Web snippets or product suggestions in order to receive additional customers; furthermore, Web search engines and recommender systems revenue is obtained from advertisements and pay-per-click. This survey paper presents a review of Web Search Engines, Ranking Algorithms, Citation Analysis and Recommender Systems. In addition, Neural networks and Deep Learning are also analyzed including their use in learning relevance and ranking. Finally, this survey paper also introduces the Random Neural Network with its practical applications.

Near infrared (NIR) spectroscopy is used in many applications to gather information about chemical composition of materials. For paper waste sorting, a small number of scores computed from NIR-spectra and assuming more or less unimodal clustered data, a pixel classifier can still be crafted by hand using knowledge about chemical properties and a reasonable amount of intuition. Additional information can be gained by visual data (VIS). However, it is not obvious what features, e.g. based on color, saturation, textured areas, are finally important for successfully separating the paper classes in feature space. Hence, a rigorous feature analysis becomes inevitable. We have chosen a generic machine-learning approach to successfully fuse NIR and VIS information. By exploiting a classification tree and a variety of additional visual features, we could increase the recognition rate to 78% for 11 classes, compared to 63% only using NIR scores. A modified feature ranking measure, which takes redundancies of features into account, allows us to analyze the importance of features and reduce them effectively. While some visual features like color saturation and hue showed to be important, some NIR scores could even be dropped. Finally, we generalize this approach to analyze raw NIR-spectra instead of score values and apply it to plastic waste sorting.

Unsupervised random-walk keyphrase extraction models mainly rely on global structural information of the word graph, with nodes representing candidate words and edges capturing the co-occurrence information between candidate words. However, integrating different types of useful information into the representation learning process to help better extract keyphrases is relatively unexplored. In this paper, we propose a random-walk method to extract keyphrases using word embeddings. Specifically, we first design a new word embedding learning model to integrate local context information of the word graph (i.e., the local word collocation patterns) with some crucial features of candidate words and edges. Then, a novel random-walk ranking model is designed to extract keyphrases by leveraging such word embeddings. Experimental results show that our approach outperforms 8 state-of-the-art unsupervised methods on two real datasets consistently for keyphrase extraction.

Multi-view data that contains the data represented in many types of features has received much attention recently. The class of method utilising non-negative matrix factorization (NMF) and manifold learning to seek the meaningful latent structure of data has been popularly used for both traditional data and multi-view data. The NMF and manifold based multi-view clustering methods focus on dealing with the challenges of manifold learning and applying manifold learning on the NMF framework. This paper provides a comprehensive review of this important class of methods on multi-view data. We conduct an extensive experiments on several datasets and raise many open problems that can be dealt with in the future so a higher clustering performance can be achieved.

Thousands of academic papers are submitted at top venues each year. Manual audits are time-consuming and laborious. And the result may be influenced by human factors. This paper investigates a modularized and attention-based recurrent convolutional network model to represent academic paper and predict aspect scores. This model treats input text as module-document hierarchies, uses attention pooling CNN and LSTM to represent text, and outputs prediction with a linear layer. Empirical results on PeerRead data show that this model give the best performance among the baseline models.

Calculation of text similarity is an essential task for the text analysis and classification. It be can based, e.g., on Jaccard, cosine or other similar measures. Such measures consider the text as a bag-of-words and, therefore, lose some syntactic and semantic features of its sentences. This article presents a different measure based on the so-called artificial sentence pattern (ASP) method. This method has been developed to analyze texts in the Polish language which has very rich inflection. Therefore, ASP has utilized syntactic and semantic rules of the Polish language. Nevertheless, we argue that it admits extensions to other languages. As a result of the analysis, we have obtained several hypernodes which contain the most important words. Each hypernode corresponds to one of the examined documents, the latter being published papers from agriculture domain written in Polish. Experimental results obtained from that set of papers have been described and discussed. Those results have been visually illustrated using graphs of hypernodes and compared with Jaccard and cosine measures.

The citation similarity measurement task is defined as determining how similar the meanings of two citations are. This task play an significant role in Natural Language Processing applications, especially in academic plagiarism detection. Yet, computing citation similarity is not a trivial task, due to the incomplete and ambiguous information presented in academic papers, which makes necessity to leverage extra knowledge to understand it, as well as most similarity measures based on the syntactic features, and other based on the semantic part still has many drawbacks. In this paper, we propose a corpus-based approach using deep learning word embeddings to compute more effective citation similarity. Our study explores the previous works on text similarity, namely, string-based, knowledge-based and corpus-based. Then we define our new basis and experiment on a large dataset of scientific papers. The final results demonstrate that deep learning based approach can enhance the effectiveness of citation similarity.

The paper addresses the task of information extraction from scientific literature with machine learning methods. In particular, the tasks of definition and result extraction from scientific publications in Russian are considered. We note that annotation of scientific texts for creation of training dataset is very labor insensitive and expensive process. To tackle this problem, we propose methods and tools based on active learning. We describe and evaluate a novel adaptive density-weighted sampling (ADWeS) meta-strategy for active learning. The experiments demonstrate that active learning can be a very efficient technique for scientific text mining, and the proposed meta-strategy can be beneficial for corpus annotation with strongly skewed class distribution. We also investigate informative task-independent features for information extraction from scientific texts and present an openly available tool for corpus annotation, which is equipped with ADWeS and compatible with well-known sampling strategies.

Electrocardiogram (ECG) is one of the most practiced methods to detect any abnormalities in human heart function. ECG waveforms are usually recorded as paper form. However, ECG paper records are inconvenient for storage and retrieval of the patient records. In this paper, an improved ECG waveform extraction algorithm from paper records is proposed. It is used to get accurate ECG trajectory information with a serious of adaptive image processing techniques, including skew correction, waveform segmentation, and tracking. The presented algorithm is tested with a number of ECG records printed from different equipment. Furthermore, three metrics are adopted for quantitative measurement between reconstructed signals and original waveforms. The comparison result shows an average accuracy of 95.5%, which proves the effectiveness of the method.

Predicting the most impactful (big-hit) paper among a researcher’s publications so it can be well disseminated in advance not only has a large impact on individual academic success, but also provides useful guidance to the research community. In this work, we tackle the problem of given the corpus of a researcher’s publications in previous few years, how to effectively predict which paper will become the big-hit in the future. We explore a series of features that can drive a paper to become the big-hit, and design a novel Bayesian ranking algorithm AdaWIRL (Adaptive Weighted Impact Ranking Learning) that leverages a weighted training schema and an adaptive timely false correction strategy to predict big-hit papers. Experimental results on the large ArnetMiner dataset with over 1.7 million authors and 2 million papers demonstrate the effectiveness of AdaWIRL. Specifically, it correctly predicts over 78.3 % of all researchers’ big-hit papers and outperforms the compared regression and ranking algorithms, with an average of \(5.8\,\%\) and \(2.9\,\%\) improvement respectively. Further analysis shows that temporal features are the best indicator for personal big-hit papers, while authorship and social features are less relevant. We also demonstrate that there is a high correlation between the impact of a researcher’s future works and their similarity to the predicted big-hit paper.

Spherical fuzzy set (SFS) is also one of the fundamental concepts for address more uncertainties in decision problems than the existing structures of fuzzy sets, and thus its implementation was more substantial. The well-known sine trigonometric function maintains the periodicity and symmetry of the origin in nature and thus satisfies the expectations of the experts over the multi parameters. Taking this feature and the significance of the SFSs into the consideration, the main objective of the article is to describe some reliable sine trigonometric laws for SFSs. Associated with these laws, we develop new average and geometric aggregation operators to aggregate the Spherical fuzzy numbers. Then, we presented a group decision-making strategy to address the multi-attribute group decision-making problem using the developed aggregation operators. To verify the value of the defined operators, a MAGDM strategy is provided along with an application for the selection of an authentic COVID-19 laboratory. Moreover, a comparative study is also performed to present the effectiveness of the developed approach.

Neighborhood systems are used to approximate graphs as finite topological structures. Throughout this article, we construct new types of eight neighborhoods for vertices of an arbitrary graph, say, j-adhesion neighborhoods. Both notions of Allam et al. and Yao will be extended via j-adhesion neighborhoods. We investigate new types of j-lower approximations and j-upper approximations for any subgraph of a given graph. Then, the accuracy of these approximations will be calculated. Moreover, a comparison between accuracy measures and boundary regions for different kinds of approximations will be discussed. To generate j-adhesion neighborhoods and rough sets on graphs, some algorithms will be introduced. Finally, a sample of a chemical example for Walczak will be introduced to illustrate our proposed methods.

Image registration is one of the most challenging problems in medical image analysis. In the recent years, deep learning based approaches became quite popular, providing fast and performing registration strategies. In this short paper, we summarise our work presented on Learn2Reg challenge 2020. The main contributions of our work rely on (i) a symmetric formulation, predicting the transformations from source to target and from target to source simultaneously, enforcing the trained representations to be similar and (ii) integration of variety of publicly available datasets used both for pretraining and for augmenting segmentation labels. Our method reports a mean dice of 0.64 for task 3 and 0.85 for task 4 on the test sets, taking third place on the challenge. Our code and models are publicly available at https://github.com/TheoEst/abdominal_registration and https://github.com/TheoEst/hippocampus_registration.

Neural networks have many applications in safety and mission critical systems. As industrial standards in various safety-critical domains require developers of critical systems to provide safety assurance, tools and techniques must be developed that enable effective creation of safety evidence for AI systems. In this position paper, we propose the use of rules extracted from neural networks as artefacts for safety evidence. We discuss the rationale behind the use of rules and illustrate it using the MNIST dataset.

Compressed air systems are often the most expensive and inefficient industrial systems. For every 10 units of energy, less than 1 unit turns into useful compressed air. Air compressors tend to be kept fully on even if they are not (all) needed. The research proposed in this short paper will combine real time ambient sensing with Artificial Intelligence and Knowledge Management to automatically improve efficiency in energy intensive manufacturing. The research will minimise energy use for air compressors based on real-time manufacturing conditions (and anticipated future requirements). Ambient data will provide detailed information on performance. Artificial Intelligence will make sense of that data and automatically act. Knowledge Management will facilitate the processing of information to advise human operators on actions to reduce energy use and maintain productivity. The aim is to create new intelligent techniques to save energy in compressed air systems.

Many companies are studying autonomous vehicles. One trend in the development of control algorithms for autonomous vehicles is the use of deep-learning approaches. The general idea is to simulate a human driver’s decision-making and behavior in various scenarios without necessarily knowing why the decision is made. In this position paper, we first argue that traditional safety analysis methods need to be extended to verify deep-learning-based autonomous vehicles. Then, we propose borrowing ideas from the process of issuing driving licenses to human drivers to verify autonomous vehicles. Verification of autonomous vehicles could focus on sufficient training as well as mental and physical health checks. Based on this position, we list several challenges that need to be addressed.

This position paper discusses the definition and implementation of Web-Physical (WP) user profiles, which allow the creation of personalized recommendations and innovative behavioral predictions in particular scenarios, i.e., fairs. The nature of a WP profile builds upon two different worlds: the Web (social networks and web applications) and the Physical one, each one of them being explored through (big) data collection platforms. These two platforms collect radically different information: on the one hand, information of appreciation towards a particular product or service (web domain) together with other metadata; on the other, the leases (x, y) of users in the exhibition space (physical domain). In this scenario, our research idea consists in identifying how the information in the two domains can be merged in a whole entity under a theoretical point of view: this will unleash tangible repercussions in terms of personalized recommendations and effective behavioral predictions, where with personalized recommendation we mean a suggestion to a user in physical terms (eg a pavilion to visit) and / or in web terms (eg a site to visit) and with behavioral prediction a prediction of where a user can go in the future, even in a multimedia perspective (physical + web).

The traditional role of nearest-neighbor classification in music classification research is that of a straw man opponent for the learning approach of the hour. Recent work in high-dimensional indexing has shown that approximate nearest-neighbor algorithms are extremely scalable, yielding results of reasonable quality from billions of high-dimensional features. With such efficient large-scale classifiers, the traditional music classification methodology of aggregating and compressing the audio features is incorrect; instead the approximate nearest-neighbor classifier should be given an extensive data collection to work with. We present a case study, using a well-known MIR classification benchmark with well-known music features, which shows that a simple nearest-neighbor classifier performs very competitively when given ample data. In this position paper, we therefore argue that nearest-neighbor classification has been treated unfairly in the literature and may be much more competitive than previously thought.

It is conventional wisdom in inductive rule learning that shorter rules should be preferred over longer rules, a principle also known as Occam’s Razor. This is typically justified with the fact that longer rules tend to be more specific and are therefore also more likely to overfit the data. In this position paper, we would like to challenge this assumption by demonstrating that variants of conventional rule learning heuristics, so-called inverted heuristics, learn longer rules that are not more specific than the shorter rules learned by conventional heuristics. Moreover, we will argue with some examples that such longer rules may in many cases be more understandable than shorter rules, again contradicting a widely held view. This is not only relevant for subgroup discovery but also for related concepts like characteristic rules, formal concept analysis, or closed itemsets.

Users’ opinions can be greatly beneficial in developing and providing products and services and improving marketing techniques for customer recommendation and retention. For this reason, sentiment analysis algorithms that automatically extract sentiment information from customers’ reviews are receiving growing attention from the computer science community. Aspect-based sentiment analysis (ABSA) allows for a more detailed understanding of customer opinions because it enables extracting sentiment polarities along with the sentiment target from sentences. ABSA consists of two steps: Aspect Extraction (AE) that allows recognizing the target sentiment; Aspect Sentiment Classification (ASC) that enables to classify the sentiment polarity. Currently, most diffused sentiment analysis algorithms are based on deep learning. Such algorithms require large labeled datasets that are extremely expensive and time consuming to build. In this paper, we present two approaches based on transfer learning and weak supervision, respectively. Both have the goal of reducing the manual effort needed to build annotated datasets for the ASC problem. In the paper, we describe the two approaches and experimentally compare them.

Community detection has become an important research direction for data mining in complex networks. It aims to identify topological structures and discover patterns in complex networks, which presents an important problem of great significance. In this paper, we are interested in the detection of communities in the Protein-Protein or Gene-gene Interaction (PPI) networks. These networks represent a set of proteins or genes that collaborate at the same cellular function. The goal is to identify such semantic and topological communities from gene annotation sources such as Gene Ontology. We propose a Genetic Algorithm (GA) based approach to detect communities having different sizes from PPI networks. For this purpose, we introduce three specific components to the GA: a fitness function based on a similarity measure and the interaction value between proteins or genes, a solution for representing a community with dynamic size and a specific mutation operator. In the computational tests carried out in this work, the introduced algorithm achieved excellent results to detect existing or even new communities from PPI networks.

Companies are increasingly measuring their products and services, resulting in a rising amount of available time series data, making techniques to extract usable information needed. One state-of-the-art technique for time series is the Matrix Profile, which has been used for various applications including motif/discord discovery, visualizations and semantic segmentation. Internally, the Matrix Profile utilizes the z-normalized Euclidean distance to compare the shape of subsequences between two series. However, when comparing subsequences that are relatively flat and contain noise, the resulting distance is high despite the visual similarity of these subsequences. This property violates some of the assumptions made by Matrix Profile based techniques, resulting in worse performance when series contain flat and noisy subsequences. By studying the properties of the z-normalized Euclidean distance, we derived a method to eliminate this effect requiring only an estimate of the standard deviation of the noise. In this paper we describe various practical properties of the z-normalized Euclidean distance and show how these can be used to correct the performance of Matrix Profile related techniques. We demonstrate our techniques using anomaly detection using a Yahoo! Webscope anomaly dataset, semantic segmentation on the PAMAP2 activity dataset and for data visualization on a UCI activity dataset, all containing real-world data, and obtain overall better results after applying our technique. Our technique is a straightforward extension of the distance calculation in the Matrix Profile and will benefit any derived technique dealing with time series containing flat and noisy subsequences.

Most public domain images are exposed to copyright infringement due to different Rights Management Information (RMI) expression system and/or out-of-date RMI. To solve this problem, this paper proposes a reliable method of integrating the RMI representation system and updating the RMI with the up-to-date information based on the most reliable data among information collected from each site through a comparative search technique for public domain images based on deep learning. Through the RMI updating method proposed in this paper, public domain image sites will be able to minimize copyright infringement, since they can provide the most up-to-date RMI and may transform different RMI representation systems into a unified system.

Fiducial marker systems have unique designs and various geometric shapes but all of them could be automatically detected with cameras. Fiducial marker system are used in such fields as augmented reality applications, medicine, space and robot-assisted tasks. A variety of applications determines criteria, which characterize qualitative properties of a marker and include such evaluation benchmarks as resilience to occlusion, distance to a marker, false positive and false negative rates, sensitivity to illumination, and others. This paper presents experimental comparison of existing ARTag, AprilTag, and CALTag systems with regard to their reliability and detection rate in occlusions of various types and intensities. Two camera types are used for experiments: inexpensive Web camera FaceCam 1000X and a high fidelity camera, which is a main vision sensor of a Russian humanoid robot AR-601M. In addition, we describe preliminary results of virtual experiments in ROS Gazebo environment. Our long term goal is to calibrate humanoid robot manipulators in real-world environment applying a pre-calibrated camera of the robot, while the presented in this paper results help selecting a most suitable marker system for further calibration procedures.

Purpose Automatic identification of interventional devices in X-ray (XR) fluoroscopy offers the potential of improved navigation during transcatheter endovascular procedures. This paper presents a prototype implementation of fully automatic 3D reconstruction of a cryo-balloon catheter during pulmonary vein isolation (PVI) procedures by deep learning approaches. Methods We employ convolutional neural networks (CNN) to automatically identify the cryo-balloon XR marker and catheter shaft in 2D fluoroscopy during PVI. Training data are generated exploiting established semiautomatic techniques, including template-matching and analytical graph building. A first network of U-net architecture uses a single grayscale XR image as input and yields the mask of the XR marker. A second network of the similar architecture is trained using the mask of the XR marker as additional input to the grayscale XR image for the segmentation of the cryo-balloon catheter shaft mask. The structures automatically identified in two 2D images with different angulations are then used to reconstruct the cryo-balloon in 3D. Results Automatic identification of the XR marker was successful in 78% of test cases and in 100% for the catheter shaft. Training of the model for prediction of the XR marker mask was successful with 3426 training samples. Incorporation of the XR marker mask as additional input for the model predicting the catheter shaft allowed to achieve good training result with only 805 training samples. The average prediction time per frame was 14.47 ms for the XR marker and 78.22 ms for the catheter shaft. Localization accuracy for the XR marker yielded on average 1.52 pixels or 0.56 mm. Conclusions In this paper, we report a novel method for automatic detection and 3D reconstruction of the cryo-balloon catheter shaft and marker from 2D fluoroscopic images. Initial evaluation yields promising results thus indicating the high potential of CNNs as alternatives to the current state-of-the-art solutions.

International audience; Relation extraction (RE) between a pair of entity mentions from text is an important and challenging task specially for open domain relations. Generally, relations are extracted based on the lexical and syntactical information at the sentence level. However, global information about known entities has not been explored yet for RE task. In this paper, we propose to extract a graph of entities from the overall corpus and to compute features on this graph that are able to capture some evidences of holding relationships between a pair of entities. The proposed features boost the RE performance significantly when these are combined with some linguistic features.

Process modeling is a complex and important task in any business process management project. Gathering information to build a process model needs effort by analysts in different ways, such as interviews and document review. However, this documentation is not always well structured and can be difficult to be understood. Thus, techniques that allow the structuring and recognition of process elements in the documentation can help in the understanding of the process and, consequently, in the modeling activity. In this context, this paper proposes an approach to recognize business process elements in natural language texts. We defined a set of 32 mapping rules to recognize business process elements in texts using natural language processing techniques and which were identified through an empirical study in texts containing descriptions of a process. Furthermore, a prototype was developed and it showed promising results. The analyses of 70 texts revealed 73.61% precision, 70.15% recall and 71.82% F-measure. Moreover, two surveys showed that 93.33% of participants agree with the mapping rules and that the approach helps the analysts in both the time spent and the effort made in the process modeling task. This paper is a reiteration and an evolution of the work presented in Ferreira et al. [1].

We present a hybrid algorithm to estimate lung nodule malignancy that combines imaging biomarkers from Radiologist's annotation with image classification of CT scans. Our algorithm employs a 3D Convolutional Neural Network (CNN) as well as a Random Forest in order to combine CT imagery with biomarker annotation and volumetric radiomic features. We analyze and compare the performance of the algorithm using only imagery, only biomarkers, combined imagery + biomarkers, combined imagery + volumetric radiomic features and finally the combination of imagery + biomarkers + volumetric features in order to classify the suspicion level of nodule malignancy. The National Cancer Institute (NCI) Lung Image Database Consortium (LIDC) IDRI dataset is used to train and evaluate the classification task. We show that the incorporation of semi-supervised learning by means of K-Nearest-Neighbors (KNN) can increase the available training sample size of the LIDC-IDRI thereby further improving the accuracy of malignancy estimation of most of the models tested although there is no significant improvement with the use of KNN semi-supervised learning if image classification with CNNs and volumetric features are combined with descriptive biomarkers. Unexpectedly, we also show that a model using image biomarkers alone is more accurate than one that combines biomarkers with volumetric radiomics, 3D CNNs, and semi-supervised learning. We discuss the possibility that this result may be influenced by cognitive bias in LIDC-IDRI because malignancy estimates were recorded by the same radiologist panel as biomarkers, as well as future work to incorporate pathology information over a subset of study participants., This paper has been submitted to the Journal of Digital Imaging (JDI 2020). The poster of this paper has received the 2nd prize for the Research Poster Award. Link: https://siim.org/page/20m_p_lung_node_malignancy

Model-driven engineering (MDE) enables to generate software tools by systematically modeling and transforming this models. However, the usability of these tools is far from efficient. Common MDE activities, such as creating a domain-specific language, are non-trivial and often require repetitive tasks. This results in unnecessary increases of development time. The goal of this paper is to increase the productivity of modelers in their every day activities by automating the tasks they perform in current MDE tools. We propose an MDE-based solution where the user defines a reusable workflow that can be parametrized at run-time and executed. Our solution works for frameworks that support two level metamodeling as well as deep metamodeling. We implemented our solution in the MDE tool AToMPM. We also performed an empirical evaluation of our approach and showed that we reduce both mechanical and thinking efforts of the user. The ideas and concepts of this paper were introduced at the MODELSWARD conference [1] and are extended in this paper.

This paper focuses on enabling the use of negotiation for complex system optimisation, whose main challenge nowadays is scalability. Although multi-agent automated negotiation has been studied for decades, it is still a challenge to handle in a scalable and efficient manner negotiation problems involving many issues with complex interdependencies. This is a clear obstacle for the use of automated negotiation in complex networks. This paper proposes a novel perspective on the negotiation process as a competitive belief propagation process, where the whole negotiation is modelled as a factor graph and distributed belief propagation techniques (BP) are used to yield a solution. We show that the model adequately suits both simple and complex negotiation settings in the literature, and we validate its efficiency and scalability in a challenging, network structured, channel negotiation setting.

Ad hoc teamwork has been introduced as a general challenge for AI and especially multiagent systems [16]. The goal is to enable autonomous agents to band together with previously unknown teammates towards a common goal: collaboration without pre-coordination. A long-term vision for ad hoc teamwork is to enable robots or other autonomous agents to exhibit the sort of flexibility and adaptability on complex tasks that people do, for example when they play games of “pick-up” basketball or soccer. As a testbed for ad hoc teamwork, autonomous robots have played in pick-up soccer games, called “drop-in player challenges”, at the international RoboCup competition. An open question is how best to evaluate ad hoc teamwork performance—how well agents are able to coordinate and collaborate with unknown teammates—of agents with different skill levels and abilities competing in drop-in player challenges. This paper presents new metrics for assessing ad hoc teamwork performance, specifically attempting to isolate an agent’s coordination and teamwork from its skill level, during drop-in player challenges. Additionally, the paper considers how to account for only a relatively small number of pick-up games being played when evaluating drop-in player challenge participants.

382 373 M30 M33 In this paper, we present a method for automatic generation of a digital resource, which connects all indirect synonyms of a dialect term to all indirect synonyms of a corresponding term in the standard language, aiming to improve the search of a digital dialect dictionary. The method uses SWRL rules defined in the Serbian WordNet ontology to identify sets of synonymous words. It also uses e-dictionaries to produce correct lemmas in standard language that users usually employ in searches. The method was applied and evaluated on verbs and a group of nouns derived from verbs (verbal nouns). We compared the results obtained by the system to those produced by humans and achieved the accuracy of 89.7%.

In Gaussian Processes a multi-output kernel is a covariance function over correlated outputs. Using a prior known relation between outputs, joint auto- and cross-covariance functions can be constructed. Realizations from these joint-covariance functions give outputs that are consistent with the prior relation. One issue with gaussian process regression is efficient inference when scaling upto large datasets. In this paper we use approximate inference techniques upon multi-output kernels enforcing relationships between outputs. Results of the proposed methodology for theoretical data and real world applications are presented. The main contribution of this paper is the application and validation of our methodology on a dataset of real aircraft flight tests, while imposing knowledge of aircraft physics into the model.

Classification of high-dimensional time series with imbalanced classes is a challenging task. For such classification tasks, the cascade classifier has been proposed. The cascade classifier tackles high-dimensionality and imbalance by splitting the classification task into several low-dimensional classification tasks and aggregating the intermediate results. Therefore the high-dimensional data set is projected onto low-dimensional subsets. But these subsets can employ unfavorable and not representative data distributions, that hamper classifiction again. Data preprocessing can overcome these problems. Small improvements in the low-dimensional data subsets of the cascade classifier lead to an improvement of the aggregated overall results. We present two data preprocessing methods, instance selection and outlier generation. Both methods are based on point distances in low-dimensional space. The instance selection method selects representative feasible examples and the outlier generation method generates artificial infeasible examples near the class boundary. In an experimental study, we analyse the precision improvement of the cascade classifier due to the presented data preprocessing methods for power production time series of a micro Combined Heat and Power plant and an artificial and complex data set. The precision increase is due to an increased selectivity of the learned decision boundaries. This paper is an extended version of [19], where we have proposed the two data preprocessing methods. In this paper we extend the analysis of both algorithms by a parameter sensitivity analysis of the distance parameters from the preprocessing methods. Both distance parameters depend on each other and have to be chosen carefully. We study the influence of these distance parameters on the classification precision of the cascade model and derive parameter fitting rules for the \(\mu \)CHP data set. The experiments yield a region of optimal parameter value combinations leading to a high classification precision.

Para-consistent logic is a non-classical logic whose foundations allow the treatment of contradictions without invalidating the conclusions. This paper presented an attempt of using Annotated Para-consistent Analysis (APA) for supporting eXplained Artificial Intelligence (XAI) on Neural-Networks. For the study case, it was presented a situation where a binary classification model was able to correctly recognize one label but not the other with the form of the principle of explosion (p\(\wedge \lnot \)p). By plotting the linear output max-min scaled data into the paraconsistent reticulate, it was possible to properly understand that the selected architectures were being able to predict. This result is early support for using APA for Neural-Network XAI, corroborating with this paper hypothesis.

This paper addresses the following predictive business process monitoring problem: Given the execution trace of an ongoing case,and given a set of traces of historical (completed) cases, predict the most likely outcome of the ongoing case. In this context, a trace refers to a sequence of events with corresponding payloads, where a payload consists of a set of attribute-value pairs. Meanwhile, an outcome refers to a label associated to completed cases, like, for example, a label indicating that a given case completed “on time” (with respect to a given desired duration) or “late”, or a label indicating that a given case led to a customer complaint or not. The paper tackles this problem via a two-phased approach. In the first phase, prefixes of historical cases are encoded using complex symbolic sequences and clustered. In the second phase, a classifier is built for each of the clusters. To predict the outcome of an ongoing case at runtime given its (uncompleted) trace, we select the closest cluster(s) to the trace in question and apply the respective classifier(s), taking into account the Euclidean distance of the trace from the center of the clusters. We consider two families of clustering algorithms – hierarchical clustering and k-medoids – and use random forests for classification. The approach was evaluated on four real-life datasets.

This paper presents an analysis of the state of the art of artificial intelligence applications in the textile industry. A review of the existing literature was performed. This article presents three methods of analyzing textile yarn. Some techniques, used in textile fabric inspection, are presented throughout this paper, as well as the use of artificial intelligence on improving the performance of productive systems using neural networks and artificial vision. The preliminary results demonstrate that the techniques covered are an asset in obtaining defects in textile fabrics at the industrial level. Taking into account the various methods of inspection and analysis of textile yarn, all present pros and cons in applicability in the textile area. In terms of advantages, all allow a better analysis of the textile yarn and defect detection with high quality, but with applicability in more complex systems. As a disadvantage, they present the fact that they do not have an already standardized algorithm that can be used, which makes its use more complex. Some possible future applications are also described.

To manage their disease, diabetic patients need to control the blood glucose level (BGL) by monitoring it and predicting its future values. This allows to avoid high or low BGL by taking recommended actions in advance. In this paper, we conduct a comparative study of two emerging deep learning techniques: Long-Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN) for one-step and multi-steps-ahead forecasting of the BGL based on Continuous Glucose Monitoring (CGM) data. The objectives are twofold: 1) Determining the best strategies of multi-steps-ahead forecasting (MSF) to fit the CNN and LSTM models respectively, and 2) Comparing the performances of the CNN and LSTM models for one-step and multi-steps prediction. Toward these objectives, we firstly conducted series of experiments of a CNN model through parameters selection to determine its best configuration. The LSTM model we used in the present study was developed and evaluated in an earlier work. Thereafter, five MSF strategies were developed and evaluated for the CNN and LSTM models using the Root-Mean-Square Error (RMSE) with an horizon of 30 min. To statistically assess the differences between the performances of CNN and LSTM models, we used the Wilcoxon statistical test. The results showed that: 1) no MSF strategy outperformed the others for both CNN and LSTM models, and 2) the proposed CNN model significantly outperformed the LSTM model for both one-step and multi-steps prediction.

This paper concerns the problem of Unsupervised Domain Adaptation (UDA) in text classification, aiming to transfer the knowledge from a source domain to a different but related target domain. Previous methods learn the discriminative feature of target domain in terms of noisy pseudo labels, which inevitably produces negative effects on training a robust model. In this paper, we propose a novel criterion Conditional Mean Discrepancy (CMD) to learn the discriminative features by matching the conditional distributions across domains. CMD embeds both the conditional distributions of source and target domains into tensor-product Hilbert space and computes Hilbert-Schmidt norm instead. We shed a new light on discriminative feature adaptation: the collective knowledge of discriminative features of different domains is naturally discovered by minimizing CMD. We propose Aligned Adaptation Networks (AAN) to learn the domain-invariant and discriminative features simultaneously based on Maximum Mean Discrepancy (MMD) and CMD. Meanwhile, to trade off between the marginal and conditional distributions, we further maximize both MMD and CMD criterions using adversarial strategy to make the features of AAN more discrepancy-invariant. To the best of our knowledge, this is the first work to definitely evaluate the shifts in the conditional distributions across domains. Experiments on cross-domain text classification demonstrate that AAN achieves better classification accuracy but less convergence time compared to the state-of-the-art deep methods.

Stock market prediction is one of the complex analysis of all time. Different expert analysts, as well as computer scientists, are working for the development of a stable and robust platform for the prediction of future stock value. The primary challenge is the nature of the movement of the daily price, which depends on various factors. To build a predictive model for the analysis of stock data and prediction is an active area of research. However, we found only a few numbers of studies performed on Korean stock market analysis, including both KOSDAQ and KOSPI companies. This study proposed a three-stage approach based on Natural Language Processing and Deep Learning techniques to analyze, comprehends the past and present market scenario, and also predict the future value of a stock. This study involves the application of natural processing techniques and deep learning techniques on around 2500 Korean companies covering KOSDAQ and KOSPI. Firstly, this paper successfully presents the current condition of the stock and overall Korean stock exchange; secondly, it recommends the potential months and weeks for investment, and finally, it predicts the future value of a stock with high accuracy. This paper may pose as a structural framework for developing a complete stock market prediction application.