ДОСЛІДЖЕННЯ ВІДКРИТИХ НАБОРІВ ДАНИХ ВЕБ-РЕСУРСІВ У КОНТЕКСТІ ЗАСТОСУВАННЯ ЇХ ДЛЯ ТЕСТУВАННЯ РЕКОМЕНДАЦІЙНИХ СИСТЕМ

Many application scenarios for image recognition require learning of deep networks from small sample sizes in the order of a few hundred samples per class. Then, avoiding overfitting is critical. Common techniques to address overfitting are transfer learning, reduction of model complexity and artificial enrichment of the available data by, e.g., data augmentation. A key idea proposed in this paper is to incorporate additional samples into the training that do not belong to the classes of the target task. This can be accomplished by formulating the original classification task as an open set classification task. While the original closed set classification task is not altered at inference time, the recast as open set classification task enables the inclusion of additional data during training. Hence, the original closed set classification task is augmented with an open set task during training. We therefore call the proposed approach open set task augmentation. In order to integrate additional task-unrelated samples into the training, we employ the entropic open set loss originally proposed for open set classification tasks and also show that similar results can be obtained with a modified sum of squared errors loss function. Learning with the proposed approach benefits from the integration of additional “unknown” samples, which are often available, e.g., from open data sets, and can then be easily integrated into the learning process. We show that this open set task augmentation can improve model performance even when these additional samples are rather few or far from the domain of the target task. The proposed approach is demonstrated on two exemplary scenarios based on subsets of the ImageNet and Food-101 data sets as well as with several network architectures and two loss functions. We further shed light on the impact of the entropic open set loss on the internal representations formed by the networks. Open set task augmentation is particularly valuable when no additional data from the target classes are available—a scenario often faced in practice.

The existing scientific documentation-based recommender systems focus on exploiting the citations and references information included in each research paper and also the lists of co-authors. In this way, it can be addressed the recommendation of related papers and even related authors. The approach we propose is original because instead of using each paper citations and co-authors, we relate each of the papers with their main research topics. This approach provides a semantic level superior to that currently used, which allows us to obtain useful results. We can use collaborative filtering recommender systems to recommend research topics related to each paper and also to recommend papers related to each research topic. In order to face this innovative proposal, we have solved a series of challenges that allow us to offer various resources and results in the paper. Our main contributions are: 1) making a data mining of scientific documentation; 2) creating and publishing an open database containing the data mining results; 3) extracting the research topics from the available scientific documentation; 4) creating and publishing a recommender system data set obtained from the database and the research topics; 5) testing the data set through a complete set of collaborative filtering methods and quality measures; and 6) selecting and showing the best methods and results, obtained using the open data set, in the context of scientific documentation recommendations. Results of the paper show the suitability of the provided data set in collaborative filtering processes, as well as the superiority of the model-based methods to face scientific documentation recommendations.

Many existing recommendation methods such as matrix factorization (MF) mainly rely on user---item rating matrix, which sometimes is not informative enough, often suffering from the cold-start problem. To solve this challenge, complementary textual relations between items are incorporated into recommender systems (RS) in this paper. Specifically, we first apply a novel weighted textual matrix factorization (WTMF) approach to compute the semantic similarities between items, then integrate the inferred item semantic relations into MF and propose a two-level matrix factorization (TLMF) model for RS. Experimental results on two open data sets not only demonstrate the superiority of TLMF model over benchmark methods, but also show the effectiveness of TLMF for solving the cold-start problem.

PurposeThis paper aims to examine the use and acceptance of open government data (OGD) of different stakeholder groups in India.Design/methodology/approachFollowing an empirical investigation among 244 respondents in India, the paper deploys path analysis via LISREL. The conceptual construction of the Unified Theory of Acceptance and Use of Technology (UTAUT) framework is used to assess the behavioral intention to use and accept OGD. Multiple regression analysis informs the relationship of demographic variables with the behavioral intention to use and accept OGD.FindingsPath analysis showed that there is an increased use and acceptance of OGD among the respondents. Multiple regression analysis shows that men are more likely to use open data sets than women. Individual differences are found among the respondents with regard to OGD usage. Thus, while men may tap open data sets in line with their purposes and professional backgrounds, women are likely to tap the data sets even for non-professional purposes. Furthermore, the respondents who are younger in age are more behaviorally inclined to accepting and using OGD than their older counterparts.Practical implicationsThe Indian government needs to popularize and familiarize OGD initiatives among the people to a greater extent. The utilitarian value of OGD may be provided when data sets are more user-friendly, frequently updated and accurate. This would facilitate in better provision of public services besides appreciating the public sentiment. Further, with increased interaction between citizens and the government, public accountability and transparency may be better realized.Social implicationsThe study shows that different groups of people are engaging in tapping information through government websites and related portals. Societal influence was found to be an important factor which predicts the acceptance and usage of OGD. An infrastructure can help to enable the use of OGD.Originality/valueHitherto, studies have focused on OGD initiatives in the developed countries, but extant literature is scarce on developing countries. Therefore, this study seeks to fill the literature gap by probing OGD use and acceptance among different stakeholder groups in India.

Open data sets that contain personal information are susceptible to adversarial attacks even when anonymized. By performing low-cost joins on multiple datasets with shared attributes, malicious users of open data portals might get access to information that violates individuals' privacy. However, open data sets are primarily published using a release-and-forget model, whereby data owners and custodians have little to no cognizance of these privacy risks. We address this critical gap by developing a visual analytic solution that enables data defenders to gain awareness about the disclosure risks in local, joinable data neighborhoods. The solution is derived through a design study with data privacy researchers, where we initially play the role of a red team and engage in an ethical data hacking exercise based on privacy attack scenarios. We use this problem and domain characterization to develop a set of visual analytic interventions as a defense mechanism and realize them in PRIVEE, a visual risk inspection workflow that acts as a proactive monitor for data defenders. PRIVEE uses a combination of risk scores and associated interactive visualizations to let data defenders explore vulnerable joins and interpret risks at multiple levels of data granularity. We demonstrate how PRIVEE can help emulate the attack strategies and diagnose disclosure risks through two case studies with data privacy experts.

ABSTRACTThe complexity of geographic modelling is increasing; hence, preparing data to drive geographic models is becoming a time-consuming and difficult task that may significantly hinder the application of such models. Meanwhile, a huge number of data sets have been shared and have become publicly accessible through the Internet. This study presents a data similarity-based approach to automatically recommend available data sets to fulfil the data requirements of geographic models. Unified description factors are adopted to provide a consistent description of public data sets and input data requirements of geographic models. Five elementary data similarities between them, specifically content, spatial coverage, temporal coverage, spatial precision, and temporal granularity similarities, are calculated. An overall similarity is estimated from aggregating the elementary data similarities. Thereafter, the candidate data for running the models are recommended in the order of overall data similarity. As a case study, the approach has been applied to recommend data from the China National Data Sharing Platform of Earth System Science to drive the population spatialization model (PSM). The approach has successfully recommended the most related data sets to run PSM. The result also suggests that the data recommendation approach can facilitate the intelligent identification of geographic data and the building of links between the open data sets.

Social Tagging Systems (STS) are very popular web application such that millions of people join the systems and actively share their contents. These enormous number of users are flooding STS with contents and tags in an unrestrained way in that threatening the capability of the system for relevant content retrieval and information sharing. Recommender Systems (RS) is a known successful method for overcome information overload problem by filtering the relevant contents over the nonrelevant contents. Besides manage folksonomy information, STS also handle social network information of its users. Both information can be used by RS to generate a good recommendation for its users. This work proposes an enhanced method for an existing hybrid recommender system, by incorporating social network information into the input of the hybrid recommender. The recommendation generation process includes Random Walk with Restart (RWR) alongside Content-Based Filtering (CBF) and Collaborative Filtering (CF) methods. Some parameters were introduced in the system to control weight contribution of each method. A comprehensive experiment with a set of a real-world open data set in two areas, social bookmark (Delicious.com) and music sharing (Last.fm) to test the proposed hybrid recommender system. The outcomes exhibit that it can give improvement compared to an existing method in terms of accuracy. The proposed hybrid achieves 24.4% more than RWR on the Delicious dataset, and 53.85% more than CBF on Lastfm dataset.

Here we present data from two studies, both of which had the purpose of investigating the potential of using electroencephalograms measured from the ear (’ear-EEG’) for sleep monitoring in a home environment. In total, 320 nights were recorded. All nights were recorded with ear-EEG, and some were also recorded using scalp-EEG and/or wristworn actigraphy. All subjects were recorded multiple times. To our knowledge, this is the most extensive open access data set available for mobile EEG development, and possibly also the best open access dataset for studying repeated sleep monitoring on individuals. We describe the details of each data set, including data quality measures, and compare the sleep scoring performance to a previously published dataset.

Remote sensing image target object detection and recognition are widely used both in military and civil fields. There are many models proposed for this purpose, but their effectiveness on target object detection in remote sensing images is not ideal due to the influence of climate conditions, obstacles and confusing objects presented in images, image clarity, and associated problems with small-target and multi-target detection and recognition. Therefore, how to accurately detect target objects in images is an urgent problem to be solved. To this end, a novel model, called YOLOv4_CE, is proposed in this paper, based on the classical YOLOv4 model with added improvements, resulting from replacing the backbone feature-extraction network with a ConvNeXt network, replacing the Complete Intersection over Union (CIoU) loss with the Efficient Intersection over Union (EIoU) loss, and adding a coordinate attention mechanism to YOLOv4 as to improve the remote sensing image detection capabilities. The results, obtained through experiments conducted on two open data sets, demonstrate that the proposed YOLOv4_CE model outperforms, in this regard, both the original YOLOv4 model and four other state-of-the-art models, namely Faster R-CNN, Gliding Vertex, Oriented R-CNN, and EfficientDet, in terms of the mean average precision ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">mAP</i> ) and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">F1 score</i> , by achieving respective values of 95.03% and 0.933 on the NWPU VHR-10 data set, and 95.89% and 0.937 on the RSOD data set.

Direct-To-Consumer (DTC) genetic testing is still a rather recent but growing phenomenon, with some companies now having as many as 850k paying customers (1). DTC genetic tests are still largely based on the analysis of Single Nucleotide Polymorphisms (SNPs) with micro arrays. The analysis of such SNP data sets in general, using Genome-Wide Association Studies (GWAS), has lead to the discovery of many genotype-phenotype associations for a wide array of phenotypic traits and disesases (2,3). The growing number of customers also enabled providers of DTC genetic testing to perform their own GWAS (4). By and large the data generated in the field of DTC genetic testing is not openly available to third parties, like academic researchers, citizen scientists, and hobbyist genealogists. This is partially due to privacy and ethical concerns (5,6), but also due to financial interests of the DTC providers (7,8). With openSNP we created an easily accessible platform to enable customers of DTC to dedicate their own DTC genetic testing data along with their phenotypes into the public domain using Creative Commons Zero. Additionally the platform offers annotations for those SNPs generated by mining different public and open data sets, such as the Public Library of Science, the SNPedia (9), Mendeley and more. Since openSNP started at the end of 2011 people have used it to release over 1700 genotyping files into the public domain and included information for over 289 phenotypes, thus creating a free and growing resource for scientists and citizen scientists alike. With this the project actively contributes to the discussion on open human genetic data, such as bioethical implications and privacy research (10,11), genealogy, teaching, pharmacogenomics (12) and even art (13). The video of this presentation is available on the BOSC YouTube channel: view video . REFERENCES (1 ) http://mediacenter.23andme.com/blog/2015/02/19/fdabloomupdate/ (2) Johnson A, O’Donnell C. BMC Med. Genet. (2009); 10: 6. (3) Hindorff LA, Sethupathy P, Junkins HA, Ramos EM, Mehta JP, et al. Proc Nat Acad Sci USA. (2009); 106: 9362-9367. (4) Do CB, Tung JY, Dorfman E, Kiefer AK, et al. PLoS Genet. (2011); 7(6): e1002141. (5) Caulfield T, McGuire AL. Annu. Rev. Med. (2011); 63: 1.1-1.11. (6) Hogarth S, Javitt G, Melzer D. Annu. Rev. Genomics Hum. Genet. (2008); 9: 161–82. (7) http://mediacenter.23andme.com/blog/2015/01/06/23andme-genentech-pd/ (8) http://mediacenter.23andme.com/blog/2015/01/12/23andme-pfizer-research-platform/ (9) Cariaso M and Lennon G, Nucl. Acids Res. (2012) 40 (D1):D1308-D1312. (10) Angrist M, PLoS ONE (2014) 9(3): e92060. (11) http://seclab.soic.indiana.edu/GenomePrivacy/papers/Genome%20Privacy-paper4.pdf (12) Samwald et al. BMC Medical Informatics and Decision Making (2015) 15:12 (13) https://soundcloud.com/thesoundofpeople/the-sound-of-bastian-greshake-6-channels-remixed

With the development of network scale, network technology affects every aspect of people&#x0027;s life. It is of great significance to detect network intrusion. Traditional research is mostly based on open data sets, the open data sets lack timeliness, and the validity of the research results is unknown. Based on the previous research, this paper proposed a novel intrusion detection method based on convolutional neural network. Firstly, real abnormal data packets were obtained by building a network environment and using real network attack tools. Second, abnormal data packets were used to generate features. Furthermore those futures are transformed into gray images for visual analysis. In order to evaluate effectiveness and superiority of proposed method, several evaluating indicators were introduced. The experimental result shows that precision, recall and F1 value of the proposed method reached 0.99, 0.99 and 0.99 respectively, which were all superior to the traditional machine learning methods.

Building archetype identification is crucial for Urban Building Energy Modeling (UBEM), but is still considered one of the biggest challenges in this field. New methods of data acquisition, along with data mining techniques such as clustering, have recently received attention for the possibility of significantly increasing identification reliability and archetype accuracy. This paper aims to establish a new and simple clustering methodology for developing building archetypes for hybrid UBEM, using open data sets and multiple diverse variables, that is still reliable and possible to validate without the use of metered energy use or real building data. The methodology uses k-means clustering for 10 building parameters simultaneously, including socio-economic parameters obtained using spatial interpolation of statistical values. Building archetypes are successfully developed for the residential building stocks of two case study areas in Sweden. The results also show that the error metric values for multiple iterations diverge after a certain number of clusters, even when using the same clustering methodology on the same data set. This discovered effect, along with the combined use of one well-known and one novel error metric, constitutes a framework well adapted to accurately determining the optimal number of building archetypes.

In this paper, we present a data set and methods to compare speech processing models and human behaviour on a phone discrimination task. We provide Perceptimatic, an open data set which consists of French and English speech stimuli, as well as the results of 91 English-and 93 French-speaking listeners. The stimuli test a wide range of French and English contrasts, and are extracted directly from corpora of natural running read speech, used for the 2017 Zero Resource Speech Challenge. We provide a method to compare humans' perceptual space with models' representational space, and we apply it to models previously submitted to the Challenge. We show that, unlike unsupervised models and supervised multilingual models, a standard supervised monolingual HMM-GMM phone recognition system, while good at discriminating phones, yields a representational space very different from that of human native listeners.

Clouds lead to missing or distorted land-related information in impacted areas in optical remote sensing images. Cloud masking, which labels cloud-contaminated pixels, forms the basis for subsequent image utilization, such as excluding the distorted pixel or filling in the missing area. However, due to the diverse spectral, textural, and shape characteristics of different clouds and complicated combinations with the underlying land surfaces, cloud masking has become a challenge in remote sensing image processing. In recent years, the Mask region-based convolutional neural network (R-CNN) method, which performs instance segmentation from a complex background and generates a pixelwise mask for the object of interest, has been used widely in object segmentation tasks. When the Mask R-CNN method is used for cloud masking, the mask result has certain problems, such as failing to extract uncommon clouds and outputting inaccurate mask boundaries for large clouds. To address these problems, we introduce two strategies, group training and boundary optimization, to improve the Mask R-CNN. For group training, samples are divided into several groups. The samples in the first group are used for the initial training, and the samples in the next group are used for evaluation. Only samples with missing or falsely detected clouds are used for tuning the classifier; then, these processes are repeated until all groups have been used or the detection precision becomes stable. For boundary optimization, a block-by-block mask strategy is adopted to guarantee that clouds with diverse sizes have similar performances. Finally, two open data sets and one data set labelled by ourselves are selected to test the proposed method, and the results demonstrate that our method can produce cloud masks for different cloud types and diverse underlying land surfaces and can achieve high accuracies, thereby providing an effective alternative for cloud masking. Compared with the original Mask R-CNN method, our method improves the average recall, average precision, and intersection over union by 5.88%, 2.4%, and 0.071 in pixel level, respectively, demonstrating the effectiveness of our improvement.

Learning the latent representations of entities based on their relationships and the data associated with them is an essential task in many applications such as ranking, recommendation systems, graph-based team formation, keyword search, and many more. However, the majority of existing techniques learn the latent representations of either network or textual data. Structural embedding techniques suffer from the sparsity of real-world networks. Attributes of nodes are a source of rich information to ameliorate network embedding vectors which are overlooked in the literature. Thus, most existing network representation learning tools capture structural information. This paper introduces an open-source toolkit called OpenAttHetRL to learn the latent representations of entities based on their both network and textual data in an end-to-end fashion. OpenAttHetRL is easy to employ and adapt for a variety of tasks including ranking, recommendation systems, and expert finding. OpenAttHetRL aims to provide a unified toolkit for data pre-processing, building and training models, and performing predictions for a downstream task. It employs a graph convolution network to capture the relationships among entities and a kernel pooling technique to preserve the similarity of their textual data in the embedding space. We use expert finding in community question answering systems to demonstrate how OpenAttHetRL can be trained to get latent representations of questions, their askers, tags, and answerers and find potential answerers of new questions.