Data Evaluation

Cross section data were evaluated for the production of the medically important positron emitter 124 I (T 1/2 ¼ 4.18 d) via 3 He-and a-particle induced reactions on Sb isotopes. The consistency in the measured data available in the literature was checked against the cross section calculations of three nuclear model codes (i.e. STAPRE, EMPIRE and TALYS). The recommended excitation functions obtained by a statistical procedure were used to derive the integral yields. An assessment of the 124 I yields and associated radioisotopic impurities suggests that the 123 Sb(a,3n) 124 I process over the energy range of E a ¼ 45-32 MeV could be of potential interest for the production of 124 I.

Cross-section data for the production of medically important radionuclide 124 I via five proton and deuteron induced reactions on enriched tellurium isotopes were evaluated. The nuclear model codes, STAPRE, EMPIRE and TALYS, were used for consistency checks of the experimental data. Recommended excitation functions were derived using a well-defined statistical procedure. Therefrom integral yields were calculated. The various production routes of 124 I were compared. Presently the 124 Te(p,n) 124 I reaction is the method of choice; however, the 125 Te(p,2n) 124 I reaction also appears to have great potential.

The computerization of scientific data treatment and the relatively recent awareness of the fragility of the natural world environment (Rio Conference "Earth Summit" in 1992) have led to the proliferation of Information Systems dedicated to biodiversity. Given the various data that they contain, they are complex applications, focused on the needs of one type of environmental scientist, closed to amateur's contributions and unable to support ethological information. Moreover, the data contained in these systems are hardly accessible to non specialists like the general public or decisionmakers. At the same time, new communication protocols, like webservices ensure a better sharing of information between applications; while immersive representations of three dimensional virtual worlds, also known as metaverses, allow a more natural assimilation of information. By putting users in a reality reproduction built from information systems, all entities can be represented in a consistent virtual environment [15]. But sharing and turning Biodiversity Information System's data in a coherent metaverse is not a trivial process. It relies on an adapted architecture and the availability of specific metadata, as several virtual worlds representing several levels of details can be generated. This paper is a contribution to the enhancement of Biodiversity Information Systems in order to make them more adaptable, usable, and representable for several kinds of users: different types of specialists, amateurs, and the general public. The MABIS modular architecture is presented as an open, efficient and evolutionary model for structuring Biodiversity Information System. The advantages of its architecture allow to ease its development, store ethological data, manage, share and represent complex entities and metadata while ensuring their authentication through an evaluation process.

Machine learning has the potential to assist many communities in using the large datasets that are becoming more and more available. Unfortunately, much of that potential is not being realized because it would require sharing data in a way that compromises privacy. In this paper, we investigate a method for ensuring (differential) privacy of the generator of the Generative Adversarial Nets (GAN) framework. The resulting model can be used for generating synthetic data on which algorithms can be trained and validated, and on which competitions can be conducted, without compromising the privacy of the original dataset. Our method modifies the Private Aggregation of Teacher Ensembles (PATE) framework and applies it to GANs. Our modified framework (which we call PATE-GAN) allows us to tightly bound the influence of any individual sample on the model, resulting in tight differential privacy guarantees and thus an improved performance over models with the same guarantees. We also look at measuring the quality of synthetic data from a new angle; we assert that for the synthetic data to be useful for machine learning researchers, the relative performance of two algorithms (trained and tested) on the synthetic dataset should be the same as their relative performance (when trained and tested) on the original dataset. Our experiments, on various datasets, demonstrate that PATE-GAN consistently outperforms the stateof-the-art method with respect to this and other notions of synthetic data quality.

The van der Schaar Lab has released an alpha version of clairvoyance, a groundbreaking package that represents the culmination of years of research, development, and real-world testing. As a uni ed, end-to-end pipeline for time-series data, clairvoyance is unmatched in its versatility and capability. While primarily developed to aid clinical research and decision support, clairvoyance can also be used to great effect in non-medical contexts, thanks to its ability to facilitate complex inference work ows in a transparent, reproducible, and ef cient manner. clairvoyance is an enormously important project for our team, and is the result of years of work across a number of different areas. It's the rst of its kind: an end-to-end pipeline that can produce personalized and interpretable predictions and recommendations using time-series data. I have no doubt that    

Machine learning has the potential to assist many communities in using the large datasets that are becoming more and more available. Unfortunately, much of that potential is not being realized because it would require sharing data in a way that compromises privacy. In this paper, we investigate a method for ensuring (differential) privacy of the generator of the Generative Adversarial Nets (GAN) framework. The resulting model can be used for generating synthetic data on which algorithms can be trained and validated, and on which competitions can be conducted, without compromising the privacy of the original dataset. Our method modifies the Private Aggregation of Teacher Ensembles (PATE) framework and applies it to GANs. Our modified framework (which we call PATE-GAN) allows us to tightly bound the influence of any individual sample on the model, resulting in tight differential privacy guarantees and thus an improved performance over models with the same guarantees. We also look at m...

Application of compound-specific stable isotope approaches often involves comparisons of isotope enrichment factors (ε). Experimental determination of ε-values is based on the Rayleigh equation, which relates the change in measured isotope ratios to the decreasing substrate fractions and is valid for closed systems. Even in well-controlled batch experiments, however, this requirement is not necessarily fulfilled, since repetitive sampling can remove a significant fraction of the analyte. For volatile compounds the need for appropriate corrections is most evident, and various methods have been proposed to account for mass removal and for volatilization into the headspace. In this study we use both synthetic and experimental data to demonstrate that the determination of ε-values according to current correction methods is prone to considerable systematic errors even in well-designed experimental setups. Application of inappropriate methods may lead to incorrect and inconsistent ε-value...

The recently reported cross-section data for the production of 82 Sr via the nat Rb(p,xn) 82 Sr process were evaluated. For the nat Rb(p,xn) 85 Sr process, cross-sections were measured experimentally over the proton energy range of 25-45 MeV, a region where very few data existed. An evaluation of the recently published data on the formation of 85 Sr was then also performed. From the recommended data curves, the integral yields of the desired radionuclide 82 Sr and the impurity 85 Sr were calculated. Yields were also determined experimentally over several energy ranges using thick nat RbCl targets. The experimental and calculated yields were found to be in agreement within 15%. These integral tests add confidence to the evaluated cross-section data. For the production of 82 Sr, an incident proton energy of 60 MeV or above is recommended; the 85 Sr impurity then corresponds to o20%.

186Re is an important radionuclide having a half-life of 3.72 d that is suitable for radioimmunotherapy. Its production in no-carrier-added form is doneviacharged particle induced reactions and the data are available in EXFOR library. We evaluated two charged particle induced reactions, namely186W(p,n)186Re and186W(d,2n)186Re. In the first case, analysis was done up to about 70 MeV but in the latter only up to about 50 MeV. A statistical procedure supported by nuclear model calculations using the codes STAPRE, EMPIRE and TALYS was used to validate and fit the data. The recommended sets of data derived together with 95% confidence limits are reported. The application of those data, particularly in the calculation of integral yields is discussed. The186W(p,n)186Re reaction on highly enriched186W is presently the method of choice for production of no-carrier-added186Re and, taking into account the radionuclidic purity, the maximum recommended proton energy is 18 MeV. The formation of t...

Cross section data were evaluated for the production of the medically important positron emitter 124 I (T 1/2 ¼ 4.18 d) via 3 He-and a-particle induced reactions on Sb isotopes. The consistency in the measured data available in the literature was checked against the cross section calculations of three nuclear model codes (i.e. STAPRE, EMPIRE and TALYS). The recommended excitation functions obtained by a statistical procedure were used to derive the integral yields. An assessment of the 124 I yields and associated radioisotopic impurities suggests that the 123 Sb(a,3n) 124 I process over the energy range of E a ¼ 45-32 MeV could be of potential interest for the production of 124 I.

Cross-section data for the production of medically important radionuclide 124 I via five proton and deuteron induced reactions on enriched tellurium isotopes were evaluated. The nuclear model codes, STAPRE, EMPIRE and TALYS, were used for consistency checks of the experimental data. Recommended excitation functions were derived using a well-defined statistical procedure. Therefrom integral yields were calculated. The various production routes of 124 I were compared. Presently the 124 Te(p,n) 124 I reaction is the method of choice; however, the 125 Te(p,2n) 124 I reaction also appears to have great potential.

The imputation of missing data in healthcare records is a critical task for ensuring the integrity and utility of medical datasets. Traditional methods often rely on simplistic approaches, leading to potential biases and inaccuracies in downstream analyses. In this paper, we propose a novel machine learning approach for imputing missing healthcare data, aiming to enhance the accuracy and robustness of imputation while preserving the underlying patterns and relationships in the data. Our approach leverages advanced machine learning techniques, including ensemble and deep learning architectures, to effectively capture complex dependencies and correlations within healthcare datasets. We demonstrate the efficacy of our method through comprehensive experiments on realworld healthcare datasets, showcasing superior imputation performance compared to conventional techniques. Furthermore, we discuss the interpretability and scalability aspects of our approach, highlighting its potential for practical deployment in healthcare analytics and decision support systems. Our proposed machine learning approach offers a promising solution for addressing missing data challenges in healthcare, paving the way for more accurate and reliable data-driven insights in medical research and practice.

The imputation of missing data in healthcare records is a critical task for ensuring the integrity and utility of medical datasets. Traditional methods often rely on simplistic approaches, leading to potential biases and inaccuracies in downstream analyses. In this paper, we propose a novel machine learning approach for imputing missing healthcare data, aiming to enhance the accuracy and robustness of imputation while preserving the underlying patterns and relationships in the data. Our approach leverages advanced machine learning techniques, including ensemble and deep learning architectures, to effectively capture complex dependencies and correlations within healthcare datasets. We demonstrate the efficacy of our method through comprehensive experiments on real-world healthcare datasets, showcasing superior imputation performance compared to conventional techniques. Furthermore, we discuss the interpretability and scalability aspects of our approach, highlighting its potential for practical deployment in healthcare analytics and decision support systems. Our proposed machine learning approach offers a promising solution for addressing missing data challenges in healthcare, paving the way for more accurate and reliable data-driven insights in medical research and practice.

Auch wenn schulische Lehrpersonen nachgewiesenermaßen im Vergleich zu anderen Berufsgruppen überdurchschnittlich gut mit digitalen Medien ausgestattet sind und diese ausgiebig für die berufliche Kommunikation und Arbeitsprozesse nutzen, scheint der mediendidaktische Einsatz im schulischen Unterricht noch ein Novum zu sein. Im Folgenden wird die Vorarbeit zur Studie Lernen durch Mediengestaltung – Entwicklung eines Konzeptes zur Unterstützung mediendidaktischer Lehre im Schulalltag (Tappe 2018) vorgestellt, in der Faktoren eruiert werden, die einen didaktischen Einsatz von digitalen Medien im Unterricht fördern oder verhindern können. Dazu wird auf das Modell der Unified Theory of Acceptance and Use of Technology (UTAUT) zurückgegriffen. Um diese aus der Akzeptanzforschung stammende Theorie in ein didaktisches Setting zu überführen, sind gewisse Anpassungen notwendig. Die ersten Ergebnisse einer daran anschließenden Fragebogenstudie werden mit Hilfe einer regressionsanalytischen Ausw...

Besides reproducing tabular data properties of standalone tables, synthetic relational databases also require modeling the relationships between related tables. In this paper, we propose the Row Conditional-Tabular Generative Adversarial Network (RC-TGAN), a novel generative adversarial network (GAN) model that extends the tabular GAN to support modeling and synthesizing relational databases. The RC-TGAN models relationship information between tables by incorporating conditional data of parent rows into the design of the child table's GAN. We further extend the RC-TGAN to model the influence that grandparent table rows may have on their grandchild rows, in order to prevent the loss of this connection when the rows of the parent table fail to transfer this relationship information. The experimental results, using eight real relational databases, show significant improvements in the quality of the synthesized relational databases when compared to the benchmark system, demonstrating the effectiveness of the RC-TGAN in preserving relationships between tables of the original database.

Besides reproducing tabular data properties of standalone tables, synthetic relational databases also require modeling the relationships between related tables. In this paper, we propose the Row Conditional-Tabular Generative Adversarial Network (RC-TGAN), a novel generative adversarial network (GAN) model that extends the tabular GAN to support modeling and synthesizing relational databases. The RC-TGAN models relationship information between tables by incorporating conditional data of parent rows into the design of the child table's GAN. We further extend the RC-TGAN to model the influence that grandparent table rows may have on their grandchild rows, in order to prevent the loss of this connection when the rows of the parent table fail to transfer this relationship information. The experimental results, using eight real relational databases, show significant improvements in the quality of the synthesized relational databases when compared to the benchmark system, demonstrating the effectiveness of the RC-TGAN in preserving relationships between tables of the original database.

Besides reproducing tabular data properties of standalone tables, synthetic relational databases also require modeling the relationships between related tables. In this paper, we propose the Row Conditional-Tabular Generative Adversarial Network (RC-TGAN), a novel generative adversarial network (GAN) model that extends the tabular GAN to support modeling and synthesizing relational databases. The RC-TGAN models relationship information between tables by incorporating conditional data of parent rows into the design of the child table's GAN. We further extend the RC-TGAN to model the influence that grandparent table rows may have on their grandchild rows, in order to prevent the loss of this connection when the rows of the parent table fail to transfer this relationship information. The experimental results, using eight real relational databases, show significant improvements in the quality of the synthesized relational databases when compared to the benchmark system, demonstrating the effectiveness of the RC-TGAN in preserving relationships between tables of the original database.

As the emphasis on evidence-based policymaking in international development increases, so too should the attention paid to the quality of the research on which this evidence is based. One way to encourage this is by archiving research data to enable reanalysis, but this requirement is often ignored or resisted by development researchers. Similarly, ambivalent feelings are expressed about revisits to former research sites to conduct further research by original and other researchers. In this article, we outline why and how researchers archive and reanalyze qualitative data and revisit research sites, and discuss the potential benefi ts and challenges of these practices for development research.

Cross section data were evaluated for the production of the medically important positron emitter 124 I (T 1/2 ¼ 4.18 d) via 3 He-and a-particle induced reactions on Sb isotopes. The consistency in the measured data available in the literature was checked against the cross section calculations of three nuclear model codes (i.e. STAPRE, EMPIRE and TALYS). The recommended excitation functions obtained by a statistical procedure were used to derive the integral yields. An assessment of the 124 I yields and associated radioisotopic impurities suggests that the 123 Sb(a,3n) 124 I process over the energy range of E a ¼ 45-32 MeV could be of potential interest for the production of 124 I.

Cross-section data for the production of medically important radionuclide 124 I via five proton and deuteron induced reactions on enriched tellurium isotopes were evaluated. The nuclear model codes, STAPRE, EMPIRE and TALYS, were used for consistency checks of the experimental data. Recommended excitation functions were derived using a well-defined statistical procedure. Therefrom integral yields were calculated. The various production routes of 124 I were compared. Presently the 124 Te(p,n) 124 I reaction is the method of choice; however, the 125 Te(p,2n) 124 I reaction also appears to have great potential.

Generative Adversarial Networks (GANs) have seen their use for generating synthetic data expand, from unstructured data like images to structured tabular data. One of the recently proposed models in the field of tabular data generation, CTGAN, demonstrated state-of-the-art performance on this task even in the presence of a high class imbalance in categorical columns or multiple modes in continuous columns. Many of the recently proposed methods have also derived ideas from CTGAN. However, training CTGAN requires a high memory footprint while dealing with high cardinality categorical columns in the dataset. In this paper, we propose MeTGAN, a memory-efficient version of CTGAN, which reduces memory usage by roughly 80%, with a minimal effect on performance. MeTGAN uses sparse linear layers to overcome the memory bottlenecks of CTGAN. We compare the performance of MeTGAN with the other models on publicly available datasets. Quality of data generation, memory requirements, and the privacy guarantees of the models are the metrics considered in this study. The goal of this paper is also to draw the attention of the research community on the issue of the computational footprint of tabular data generation methods to enable them on larger datasets especially ones with high cardinality categorical variables.

This paper introduces a novel generative adversarial network (GAN) for synthesizing large-scale tabular databases which contain various features such as continuous, discrete, and binary. Technically, our GAN belongs to the category of class-conditioned generative models with a predefined conditional vector. However, we propose a new formulation for deriving such a vector incorporating both binary and discrete features simultaneously. We refer to this noble definition as compound conditional vector and employ it for training the generator network. The core architecture of this network is a threelayered deep residual neural network with skip connections. For improving the stability of such complex architecture, we present a regularization scheme towards limiting unprecedented variations on its weight vectors during training. This regularization approach is quite compatible with the nature of adversarial training and it is not computationally prohibitive in runtime. Furthermore, we constantly monitor the variation of the weight vectors for identifying any potential instabilities or irregularities to measure the strength of our proposed regularizer. Toward this end, we also develop a new metric for tracking sudden perturbation on the weight vectors using the singular value decomposition theory. Finally, we evaluate the performance of our proposed synthesis approach on six benchmarking tabular databases, namely Adult, Census, HCDR, Cabs, News, and King. The achieved results corroborate that for the majority of the cases, our proposed RccGAN outperforms other conventional and modern generative models in terms of accuracy, stability, and reliability 1 .

Qualitative secondary analysis (QSA) is a method that has been applied in other disciplines even though it has rarely been explicitly used or discussed in information science. This paper discusses the epistemological and ethical issues surrounding QSA, explains the value of the method for information science research, discusses its benefits and challenges, and provides an example case study.

We measure the two dimensional instantaneous velocity fields inside an axial turbine by using the technique PIV (Particle Image Velocimetry). The measured plane is located just behind the rotor wheel and it is oriented in axial × tangential direction. We measured a single overloaded regime as well as the nominal one at tip radius. We discuss several technical difficulties appearing during this measurement, mainly with the seeding particles and with the optical access into a massive steel body. The glass dividing the channel past rotor and the empty space of diffusor has been dirtied by large oil droplets condensing at the blades and splashed to this window. We "solved" this trouble by removing the glass, because it anyway divides volumes with same static pressure. Best quality result was obtained without this glass, but its removal changes the geometry. In our contribution we focus to the effect of such rude change to the trustworthiness of the results.

Vorwort "Super-Scoring?"-das Fragezeichen im Titel dieser Publikation ist von großer Bedeutung. Es steht für die vielen offenen Fragen im Zusammenhang mit der Einführung von Punktesystemen und Skalen zur Bewertung menschlichen Verhaltens in der digitalen Welt. Ethische, rechtliche, politische, soziologische Fragen richten sich auf die Diskurse, Logiken, Machtinteressen und zuletzt auch auf das zugrundeliegende Menschenbild dieser sozial und digital vermittelten Steuerungsinstrumente. Die Auseinandersetzung mit diesen Fragen ist nicht nur mit Blick auf China, sondern auch in westlichen Demokratien hochaktuell. Das Projekt "Super-Scoring?" wurde durch das Grimme-Forschungskolleg an der Universität zu Köln mit Mitteln des Landes Nordrhein-Westfalen im Jahr 2019 gefördert. Seit Ende 2014 setzt sich das Grimme-Forschungskolleg beobachtend, forschend und vermittelnd mit aktuellen Fragen digitaler Kommunikationskulturen und gesellschaftlicher Wandlungsprozesse auseinander. Es wird gemeinsam von der Universität zu Köln und dem Grimme-Institut getragen. Am 11. Oktober 2019 luden Prof. Dr. Björn Ahl (China Studien) und Prof. Dr. Stephan Packard (Institut für Medienkultur und Theater) von der Universität zu Köln, Arne Busse von der Bundeszentrale für politische Bildung und Dr. Harald Gapski (Grimme-Institut) zu der interdisziplinär und international besetzten Fachtagung "Data-driven societal technologies in China and Western-style democracies as a new challenge for education" in Köln ein. Vorbereitend und begleitend bauten Monika Elias und Thomas Tekster (Grimme-Institut) ein thematisches Webangebot unter www.superscoring.de auf. 11 Danksagung Die Herausgeber danken allen Expert(inn)en, Referent(inn)en, Unterstützer(inne)n, Übersetzer(inn)n und Mitarbeiter(inne)n, die das Gesamtprojekt "Super-Scoring?" mit ihren jeweilen Beiträgen unterstützt und damit auch diese Publikation ermöglicht haben:

We measure the two dimensional instantaneous velocity fields inside an axial turbine by using the technique PIV (Particle Image Velocimetry). The measured plane is located just behind the rotor wheel and it is oriented in axial × tangential direction. We measured a single overloaded regime as well as the nominal one at tip radius. We discuss several technical difficulties appearing during this measurement, mainly with the seeding particles and with the optical access into a massive steel body. The glass dividing the channel past rotor and the empty space of diffusor has been dirtied by large oil droplets condensing at the blades and splashed to this window. We "solved" this trouble by removing the glass, because it anyway divides volumes with same static pressure. Best quality result was obtained without this glass, but its removal changes the geometry. In our contribution we focus to the effect of such rude change to the trustworthiness of the results.

Der Beitrag knüpft an eine multimethodologische Untersuchung zum medialen Erleben und der medialen Praxis von Kindern im vorschulischen Alter an. Theoretischer Ausgangspunkt sind (noch) marginale empirische Befunde um mediale Lebenswelten der Kinder von vier bis sechs Jahren, welche explizit die Sichtweise der kindlichen Akteurinnen und Akteure einbindet und nicht über, sondern mit Kindern forscht. Aufgrund des erschwerten methodischen Zugriffs bei dieser Altersgruppe setzte die Studie mit den Puppet Interviews eine spezifisch kindgerechte Methode ein, die in dem vorliegenden Beitrag in ihrer Um-setzung und Weiterführung kritisch beleuchtet und diskutiert wird.

Deep Generative models(DGM) are neural networks which tries to estimate the likelihood of each observation and to create new samples from the underlying distribution of data, Synthetic data generation is a specific application of deep generative models where the goal is to create realistic and novel data samples. GANs, in particular, consist of a generator network that produces new samples and a discriminator network that distinguishes between real and generated samples. Through an adversarial training process, both networks improve iteratively, with the generator learning to produce more realistic data. However, challenges exist. Ensuring diversity and realism in generated samples is an ongoing research focus. Issues like training stability, mode collapse, and addressing biases in generated data are important considerations. Ethical concerns regarding the potential misuse of synthetic data and privacy implications must be taken into account. The Triple GAN focuses on classification and class-conditional generation of data samples and also a better framework to work with as it requires less number of epochs to train, can perform well even without data argumentation and can perform semi-supervised learning(SSL) tasks better. v

Deep Generative models(DGM) are neural networks which tries to estimate the likelihood of each observation and to create new samples from the underlying distribution of data, Synthetic data generation is a specific application of deep generative models where the goal is to create realistic and novel data samples. GANs, in particular, consist of a generator network that produces new samples and a discriminator network that distinguishes between real and generated samples. Through an adversarial training process, both networks improve iteratively, with the generator learning to produce more realistic data. However, challenges exist. Ensuring diversity and realism in generated samples is an ongoing research focus. Issues like training stability, mode collapse, and addressing biases in generated data are important considerations. Ethical concerns regarding the potential misuse of synthetic data and privacy implications must be taken into account. The Triple GAN focuses on classification and class-conditional generation of data samples and also a better framework to work with as it requires less number of epochs to train, can perform well even without data argumentation and can perform semi-supervised learning(SSL) tasks better. v

The extensive usage of online applications and social media has raised serious concerns from the public regarding the exposure of their personal information. So, there is a strong need for data anonymization to prevent privacy breaches and leakages. The era of attacks on databases and servers is an old trend. Now, most attacks are based on earning access to users’ private data. There are techniques like k-anonymity and l-diversity to protect Personally Identifiable Information (PII) from adversaries. However, these techniques still cannot provide security from homogeneity attacks, and their application is limited to structural data only. Till now, the frameworks are only available to anonymize the human face data in image format. In this paper, we proposed a new architecture for protecting privacy-related information in images of Indian vehicle number plates. We propose an architecture for anonymizing the vehicle number plates using Wasserstein’s Generative Adversarial Network (WGAN...

Lehramtsstudierende müssen in den Hochschulen auf die neuen Herausforderungen der Digitalisierung vorbereitet werden. Einen Sachverhalt für Kommilitonen/Schüler verständlich aufzubereiten, muss eingeübt werden. Eine Möglichkeit stellen hierbei Erklärvideos dar. Diese werden mit digitalen Medien aufgezeichnet und auf einer Lernplattform veröffentlicht. Der Artikel widmet sich ausführlich dieser Methode. (DIPF/bc)

A major open research challenge is developing privacy-preserving machine learning methods that both achieve high performance and privacy guarantees even though the original training data contains sensitive personal information. The applications are abundant, from making cities safer, via on-demand public transportation systems to improved medical diagnosis. The goal of our PhD projects is to develop new machine learning methods for creating synthetic spatio-temporal trajectory data sets preserving the privacy of the individuals in the original data. We will 1) extend generative adversarial network (GAN) methods to learn generative spatio-temporal trajectory models and 2) develop new Bayesian Optimization methods for creating tailored privacy-preserving synthetic data sets using these generative models.

Synthetic tabular data generation becomes crucial when real data is limited, expensive to collect, or simply cannot be used due to privacy concerns. However, producing good quality synthetic data is challenging. Several probabilistic, statistical, and generative adversarial networks (GANs) based approaches have been presented for synthetic tabular data generation. Once generated, evaluating the quality of the synthetic data is quite challenging. Some of the traditional metrics have been used in the literature but there is lack of a common, robust, and single metric. This makes it difficult to properly compare the effectiveness of different synthetic tabular data generation methods. In this paper we propose a new universal metric, TabSynDex, for robust evaluation of synthetic data. TabSynDex assesses the similarity of synthetic data with real data through different component scores which evaluate the characteristics that are desirable for "high quality" synthetic data. Being a single score metric, TabSynDex can also be used to observe and evaluate the training of neural network based approaches. This would help in obtaining insights that was not possible earlier. Further, we present several baseline models for comparative analysis of the proposed evaluation metric with existing generative models.

Missing values are common in real-world datasets and pose a significant challenge to the performance of statistical and machine learning models. Generally, missing values are imputed using statistical methods, such as the mean, median, mode, or machine learning approaches. These approaches are limited to either numerical or categorical data. Imputation in mixed datasets that contain both numerical and categorical attributes is challenging and has received little attention. Machine learning-based imputation algorithms usually require a large amount of training data. However, obtaining such data is difficult. Furthermore, no considerate work has been conducted in the literature that focuses on the effects of the training and testing size with increasing amounts of missing data. To address this gap, we proposed that increasing the amount of training data will improve imputation performance. We first used generative adversarial network (GAN) methods to increase the amount of training data. We considered two state-ofthe-art GANs (tabular and conditional tabular) to add synthetic samples using observed data with different synthetic sample ratios. We then used three state-of-the-art imputation models that can handle mixed data: MissForest, multivariate imputation by chained equations, and denoising auto encoder (DAE). We proposed robust experimental setups on four publicly available datasets with different training-testing data divisions that have increasing missingness ratios. Extensive experimental results show that incorporating synthetic samples with training data achieves better performance compared to the baseline methods for mixed data imputation in both categorical and numerical variables, especially for large missingness ratios. INDEX TERMS Mixed data imputation, missing data, GANs, miss forest, MICE, denoising auto encoders.

Gaining insight into financial transactions is a legal requirement for banks and financial service providers. Unfortunately, for most data available to banks and service provider, the truth about the transaction purpose is unknown and illicit activities have a low base rate. Additionally, after reporting suspicious activity, banks will no hearing back from investigations. Consequently, many supervised approaches, which inherently rely on the ground truth, cannot be used. In the following we study the use of Generative Adversarial Networks (GAN) for data augmentation to increase the scarce amount of known fraudulent cases, and for data imputation to improve the quality of samples.

Cross section data were evaluated for the production of the medically important positron emitter 124 I (T 1/2 ¼ 4.18 d) via 3 He-and a-particle induced reactions on Sb isotopes. The consistency in the measured data available in the literature was checked against the cross section calculations of three nuclear model codes (i.e. STAPRE, EMPIRE and TALYS). The recommended excitation functions obtained by a statistical procedure were used to derive the integral yields. An assessment of the 124 I yields and associated radioisotopic impurities suggests that the 123 Sb(a,3n) 124 I process over the energy range of E a ¼ 45-32 MeV could be of potential interest for the production of 124 I.

The aim of this contribution is to present Qualitative Content Analysis as an analytical method for empirical studies in the context of educational research and to discuss the investigative potential of this method. First, theoretical background and main analytical techniques of Qualitative Content Analysis are explained and illustrated based on examples from different empirical studies in educational research. Furthermore, Qualitative Content Analysis is discussed with respect to the combination of qualitative and quantitative research methods in the sense of triangulation and mixed methods approaches. Finally, content analytical criteria of scientific quality and software for qualitative analysis will be addressed. (DIPF/Orig.)

; we would like to thank the audience for useful comments and suggestions. In addition, we would like to thank all our informants, as well as colleagues who helped us get in contact with them: Kathleen O'Connor, Joseph Tyler, and above all Doron Cohen from the University of Manchester. Special thanks also go to Erin Myers and Aaron Hutson for their help with the recording of the stimuli. Finally, we thank the anonymous reviewers of the article for their insightful comments, and Kathleen O'Connor for her proofreading. Magnitude Estimation: can it do something for your pragmatics?

The van der Schaar Lab has released an alpha version of clairvoyance, a ground-breaking package that represents the culmination of years of research, development, and real-world testing. As a unified, end-to-end pipeline for time-series data, clairvoyance is unmatched in its versatility and capability. While primarily developed to aid clinical research and decision support, clairvoyance can also be used to great effect in non-medical contexts, thanks to its ability to facilitate complex inference workflows in a transparent, reproducible, and efficient manner.

Most of the targeted discoveries in tuberculosis research have covered previously explored chemical structures but neglected physiochemical properties. Until now, no efficient prediction tools have been developed to discriminate the novelty of screened compounds at early stages. To overcome this deficit, a drastic novel approach must include physicochemical properties filters provided by Chemical Global Positioning System-Natural Product analysis (ChemGPS-NP). Three different screening schemes GSK, GVKBio, and NIAID provided 776, 2880, and 3779 compounds respectively and were evaluated based on their physicochemical properties and thereby proposed as deduction examples. Charting the physiochemical property spaces of these sets identified the merits and demerits of each screening scheme by simply observing the distribution over the chemical property space. We found that GSK screening set was confined to a certain space, losing potentially active compounds when compared with an in-hou...

186Re is an important radionuclide having a half-life of 3.72 d that is suitable for radioimmunotherapy. Its production in no-carrier-added form is doneviacharged particle induced reactions and the data are available in EXFOR library. We evaluated two charged particle induced reactions, namely186W(p,n)186Re and186W(d,2n)186Re. In the first case, analysis was done up to about 70 MeV but in the latter only up to about 50 MeV. A statistical procedure supported by nuclear model calculations using the codes STAPRE, EMPIRE and TALYS was used to validate and fit the data. The recommended sets of data derived together with 95% confidence limits are reported. The application of those data, particularly in the calculation of integral yields is discussed. The186W(p,n)186Re reaction on highly enriched186W is presently the method of choice for production of no-carrier-added186Re and, taking into account the radionuclidic purity, the maximum recommended proton energy is 18 MeV. The formation of t...

Generative Adversarial Networks (GANs) have seen their use for generating synthetic data expand, from unstructured data like images to structured tabular data. One of the recently proposed models in the field of tabular data generation, CTGAN, demonstrated state-of-the-art performance on this task even in the presence of a high class imbalance in categorical columns or multiple modes in continuous columns. Many of the recently proposed methods have also derived ideas from CTGAN. However, training CTGAN requires a high memory footprint while dealing with high cardinality categorical columns in the dataset. In this paper, we propose MeTGAN, a memory-efficient version of CTGAN, which reduces memory usage by roughly 80%, with a minimal effect on performance. MeTGAN uses sparse linear layers to overcome the memory bottlenecks of CTGAN. We compare the performance of MeTGAN with the other models on publicly available datasets. Quality of data generation, memory requirements, and the privacy guarantees of the models are the metrics considered in this study. The goal of this paper is also to draw the attention of the research community on the issue of the computational footprint of tabular data generation methods to enable them on larger datasets especially ones with high cardinality categorical variables.

186Re is an important radionuclide having a half-life of 3.72 d that is suitable for radioimmunotherapy. Its production in no-carrier-added form is doneviacharged particle induced reactions and the data are available in EXFOR library. We evaluated two charged particle induced reactions, namely186W(p,n)186Re and186W(d,2n)186Re. In the first case, analysis was done up to about 70 MeV but in the latter only up to about 50 MeV. A statistical procedure supported by nuclear model calculations using the codes STAPRE, EMPIRE and TALYS was used to validate and fit the data. The recommended sets of data derived together with 95% confidence limits are reported. The application of those data, particularly in the calculation of integral yields is discussed. The186W(p,n)186Re reaction on highly enriched186W is presently the method of choice for production of no-carrier-added186Re and, taking into account the radionuclidic purity, the maximum recommended proton energy is 18 MeV. The formation of t...

Accurately detecting and responding to student affect is a critical capability for adaptive learning environments. Recent years have seen growing interest in modeling student affect with multimodal sensor data. A key challenge in multimodal affect detection is dealing with data loss due to noisy, missing, or invalid multimodal features. Because multimodal affect detection often requires large quantities of data, data loss can have a strong, adverse impact on affect detector performance. To address this issue, we present a multimodal data imputation framework that utilizes conditional generative models to automatically impute posture and interaction log data from student interactions with a game-based learning environment for emergency medical training. We investigate two generative models, a Conditional Generative Adversarial Network (C-GAN) and a Conditional Variational Autoencoder (C-VAE), that are trained using a modality that has undergone varying levels of artificial data maski...

Qualitative secondary analysis (QSA) is a method that has been used in other disciplines and debated from epistemological and ethical perspectives. It has rarely been explicitly used or discussed in information science. This paper explains value of the method for  information science research, discussing its benefits and  challenges and analyzing our recent QSA study.

Many electroencephalography (EEG) applications rely on channel selection methods to remove the least informative channels, e.g., to reduce the amount of electrodes to be mounted, to decrease the computational load, or to reduce overfitting effects and improve performance. Wrapper-based channel selection methods aim to match the channel selection step to the target model, yet they require to re-train the model multiple times on different candidate channel subsets, which often leads to an unacceptably high computational cost, especially when said model is a (deep) neural network. To alleviate this, we propose a framework to embed the EEG channel selection in the neural network itself to jointly learn the network weights and optimal channels in an end-to-end manner by traditional backpropagation algorithms. We deal with the discrete nature of this new optimization problem by employing continuous relaxations of the discrete channel selection parameters based on the Gumbel-softmax trick. We also propose a regularization method that discourages selecting channels more than once. This generic approach is evaluated on two different EEG tasks: motor imagery brain-computer interfaces and auditory attention decoding. The results demonstrate that our framework is generally applicable, while being competitive with state-of-the art EEG channel selection methods, tailored to these tasks.

In diesem Artikel erortere ich eine Reihe ethischer Fragen im Zusammenhang mit dem von den USA in Auftrag gegebenen Belmont-Bericht (NATIONAL COMMISSION FOR THE PROTECTION OF HUMAN SUBJECTS OF BIOMEDICAL AND BEHAVIORAL RESEARCH 1979), wobei ich als Ausgangspunkt eine postqualitative Untersuchung mit zehn (schwarzen) Kindern im Alter von 14 bis 15 Jahren in einer sudafrikanischen Schule nutze. Ich bat die Kinder, Gruppen zu bilden, um gemeinsam uber die mogliche Relevanz bestimmter, in Australien entwickelter Szenarien zu Klimawandel fur Sudafrika nachzudenken. Die "Szenario-Aktivitat" sollte das aktive gemeinsame Lernen der Teilnehmer_innen anregen. Sie sollte auch bewusst "performativ" sein, da die in den vorgestellten Szenarien verwendeten Worte Einfluss auf die (gemeinsamen) Uberlegungen der Kinder haben wurden, fur die ich eine gewisse Verantwortung ubernommen hatte. Vor dem Hintergrund dieser Untersuchung und mit Blick auf laufende ethische Debatten uber den...