Search

Ageing is a multi-factorial physiological process and the development of novel IoT systems, tools and devices, specifically targeted to older people, must be based on a holistic framework built on robust scientific knowledge in different health domains. Furthermore, interoperability must be guaranteed using standardized frameworks or approaches. These aspects still largely lack in the specific literature. The main aim of the paper is to develop a new ontology (the NESTORE ontology) to extend the available ontologies provided by universAAL-IoT (uAAL-IoT). The ontology is based on a multidomain healthy ageing holistic model, structuring well-assessed scientific knowledge, specifically targeted to healthy older adults aged between 65 and 75. The tool is intended to support, and standardize heterogeneous data about ageing in compliance with the uAAL-IoT framework. The NESTORE ontology covers all the relevant concepts to represent 3 significant domains of ageing: (1) Physiological Status and Physical Activity Behaviour; (2) Nutrition; and (3) Cognitive and Mental Status and Social Behaviour. In total, 12 sub-ontologies were modelled with more than 60 classes and sub-classes referenced among them by using more than 100 relations and around 20 enumerations. The proposed ontology increases the uAAL collection by 40%. NESTORE ontology provides innovation both in terms of semantic content and technological approach. The thorough use of this ontology can support the development of a decision support system, to promote healthy ageing, with the capacity to do dynamic multi-scale modelling of user-specific data based on the semantic annotations of users’ profile.

This systematic review provides an overview and analysis of the survey instruments measuring technology integration (TI) in educational settings from 2010 to 2021. Given the importance of addressing aspects related to the quality of TI (e.g., the use of technology to support pedagogical goals), we paid particular attention to them. Search results from the PsycINFO, ERIC, Web of Science, and Scopus databases yielded 695 records. Thirty-five different survey instruments used in 36 studies met our eligibility criteria and were then analyzed by applyingcontent analysis. Our results indicate that the diversity of operationalizations is very high and that several instruments have no explicit conceptual or theoretical underpinnings. Most of the instruments measure aspects related to classroom practices and measure TI from the teachers' point of view. The most frequently measured pedagogical aspects of TI in instructional practices include the use of technology to (1) enhance students’ cognitive engagement, (2) promote collaboration between students, and (3) allow students to conduct research online. The study concludes with some perspectives for future research, an attempt to formulate a definition of TI, and some more general recommendations to ensure terminological unambiguity in TI research.

Anomaly Detection systems based on Machine and Deep learning are the most promising solutions to detect cyberattacks in the industry. However, these techniques are vulnerable to adversarial attacks that downgrade prediction performance. Several techniques have been proposed to measure the robustness of Anomaly Detection in the literature. However, they do not consider that, although a small perturbation in an anomalous sample belonging to an attack, i.e., Denial of Service, could cause it to be misclassified as normal while retaining its ability to damage, an excessive perturbation might also transform it into a truly normal sample, with no real impact on the industrial system. This paper presents a methodology to calculate the robustness of Anomaly Detection models in industrial scenarios. The methodology comprises four steps and uses a set of additional models called support models to determine if an adversarial sample remains anomalous. We carried out the validation using the Tennessee Eastman process, a simulated testbed of a chemical process. In such a scenario, we applied the methodology to both a Long-Short Term Memory (LSTM) neural network and 1-dimensional Convolutional Neural Network (1D-CNN) focused on detecting anomalies produced by different cyberattacks. The experiments showed that 1D-CNN is significantly more robust than LSTM for our testbed. Specifically, a perturbation of 60% (empirical robustness of 0.6) of the original sample is needed to generate adversarial samples for LSTM, whereas in 1D-CNN the perturbation required increases up to 111% (empirical robustness of 1.11).

Statistical shape modeling aims at capturing shape variations of an anatomical structure that occur within a given population. Shape models are employed in many tasks, such as shape reconstruction and image segmentation, but also shape generation and classification. Existing shape priors either require dense correspondence between training examples or lack robustness and topological guarantees. We present FlowSSM, a novel shape modeling approach that learns shape variability without requiring dense correspondence between training instances. It relies on a hierarchy of continuous deformation flows, which are parametrized by a neural network. Our model outperforms state-of-the-art methods in providing an expressive and robust shape prior for distal femur and liver. We show that the emerging latent representation is discriminative by separating healthy from pathological shapes. Ultimately, we demonstrate its effectiveness on two shape reconstruction tasks from partial data. Our source code is publicly available (https://github.com/davecasp/flowssm)., accepted for MICCAI 2022

This paper presents an overview of the second edition of HIPE (Identifying Historical People, Places and other Entities), a shared task on named entity recognition and linking in multilingual historical documents. Following the success of the first CLEF-HIPE-2020 evaluation lab, HIPE-2022 confronts systems with the challenges of dealing with more languages, learning domain-specific entities, and adapting to diverse annotation tag sets. This shared task is part of the ongoing efforts of the natural language processing and digital humanities communities to adapt and develop appropriate technologies to efficiently retrieve and explore information from historical texts. On such material, however, named entity processing techniques face the challenges of domain heterogeneity, input noisiness, dynamics of language, and lack of resources. In this context, the main objective of HIPE-2022, run as an evaluation lab of the CLEF 2022 conference, is to gain new insights into the transferability of named entity processing approaches across languages, time periods, document types, and annotation tag sets. Tasks, corpora, and results of participating teams are presented. Compared to the condensed overview, this paper contains more refined statistics on the datasets, a break down of the results per type of entity, and a discussion of the `challenges' proposed in the shared task. For code and data, see the HIPE-eval github organisation: https://github.com/hipe-eval

Device fingerprinting combined with Machine and Deep Learning (ML/DL) report promising performance when detecting cyberattacks targeting data managed by resource-constrained spectrum sensors. However, the amount of data needed to train models and the privacy concerns of such scenarios limit the applicability of centralized ML/DL-based approaches. Federated learning (FL) addresses these limitations by creating federated and privacy-preserving models. However, FL is vulnerable to malicious participants, and the impact of adversarial attacks on federated models detecting spectrum sensing data falsification (SSDF) attacks on spectrum sensors has not been studied. To address this challenge, the first contribution of this work is the creation of a novel dataset suitable for FL and modeling the behavior (usage of CPU, memory, or file system, among others) of resource-constrained spectrum sensors affected by different SSDF attacks. The second contribution is a pool of experiments analyzing and comparing the robustness of federated models according to i) three families of spectrum sensors, ii) eight SSDF attacks, iii) four scenarios dealing with unsupervised (anomaly detection) and supervised (binary classification) federated models, iv) up to 33% of malicious participants implementing data and model poisoning attacks, and v) four aggregation functions acting as anti-adversarial mechanisms to increase the models robustness.

In the last years we have witnessed the fields of geosciences and remote sensing and artificial intelligence to become closer. Thanks to both the massive availability of observational data, improved simulations, and algorithmic advances, these disciplines have found common objectives and challenges to advance the modeling and understanding of the Earth system. Despite such great opportunities, we also observed a worrying tendency to remain in disciplinary comfort zones applying recent advances from artificial intelligence on well resolved remote sensing problems. Here we take a position on research directions where we think the interface between these fields will have the most impact and become potential game changers. In our declared agenda for AI on Earth sciences, we aim to inspire researchers, especially the younger generations, to tackle these challenges for a real advance of remote sensing and the geosciences., Comment: In press at IEEE Geoscience and Remote Sensing Magazine

In this work, we propose a novel data augmentation method for clinical audio datasets based on a conditional Wasserstein Generative Adversarial Network with Gradient Penalty (cWGAN-GP), operating on log-mel spectrograms. To validate our method, we created a clinical audio dataset which was recorded in a real-world operating room during Total Hip Arthroplasty (THA) procedures and contains typical sounds which resemble the different phases of the intervention. We demonstrate the capability of the proposed method to generate realistic class-conditioned samples from the dataset distribution and show that training with the generated augmented samples outperforms classical audio augmentation methods in terms of classification performance. The performance was evaluated using a ResNet-18 classifier which shows a mean Macro F1-score improvement of 1.70% in a 5-fold cross validation experiment using the proposed augmentation method. Because clinical data is often expensive to acquire, the development of realistic and high-quality data augmentation methods is crucial to improve the robustness and generalization capabilities of learning-based algorithms which is especially important for safety-critical medical applications. Therefore, the proposed data augmentation method is an important step towards improving the data bottleneck for clinical audio-based machine learning systems.

Computing path queries such as the shortest path in public transport networks is challenging because the path costs between nodes change over time. A reachability query from a node at a given start time on such a network retrieves all points of interest (POIs) that are reachable within a given cost budget. Reachability queries are essential building blocks in many applications, for example, group recommendations, ranking spatial queries, or geomarketing. We propose an efficient solution for reachability queries in public transport networks. Currently, there are two options to solve reachability queries. (1) Execute a modified version of Dijkstra’s algorithm that supports time-dependent edge traversal costs; this solution is slow since it must expand edge by edge and does not use an index. (2) Issue a separate path query for each single POI, i.e., a single reachability query requires answering many path queries. None of these solutions scales to large networks with many POIs. We propose a novel and lightweight reachability index. The key idea is to partition the network into cells. Then, in contrast to other approaches, we expand the network cell by cell. Empirical evaluations on synthetic and real-world networks confirm the efficiency and the effectiveness of our index-based reachability query solution.

Case and knowledge management systems are spread at the frontline across public agencies. However, such systems are dedicated for the collaboration within the agency rather than for the face-to-face interaction with the clients. If used as a collaborative resource at the frontline, case and knowledge management systems might disturb the service provision by displaying unfiltered internal information, disclosing private data of other clients, or revealing the limits of frontline employees’ competence (if they cannot explain something) or their authority (if they cannot override something). Observation in the German Public Employment Agency shows that employment consultants make use of various coping strategies during face-to-face consultations to extend existing boundaries set by the case and knowledge management systems and by the rules considering their usage. The analysis of these coping strategies unveils the forces that shape the conduct of employment consultants during their contacts with clients: the consultants’ own understanding of work, the actual and the perceived needs of the clients, and the political mission as well as the internal rules of the employment agency. The findings form a twofold contribution: First, they contribute to the discourse on work in employment agencies by illustrating how the complexities of social welfare apparatus demonstrate themselves in singular behavioural patterns. Second, they contribute to the discourse on screen-level bureaucracy by depicting the consultants as active and conscious mediators rather than passive interfaces between the system and the client.

We propose a simple new aggregation strategy for federated learning that won the MICCAI Federated Tumor Segmentation Challenge 2021 (FETS), the first ever challenge on Federated Learning in the Machine Learning community. Our method addresses the problem of how to aggregate multiple models that were trained on different data sets. Conceptually, we propose a new way to choose the weights when averaging the different models, thereby extending the current state of the art (FedAvg). Empirical validation demonstrates that our approach reaches a notable improvement in segmentation performance compared to FedAvg.

Open Educational Resources (OERs) have been widely promoted in the higher education community in recent years. However, the accessibility of OERs for people with disabilities has received relatively little attention. This paper presents the results of interviews carried out with people at higher education institutions worldwide who are involved in the creation and implementation of OERs. The goal is to gauge the awareness of accessibility issues in OERs. This paper raises the following research questions: How much do OER creators know about accessibility? What measures are needed to ensure accessibility in OERs? Results suggest that OER creators are aware about some issues around accessibility, but they still need further training on how to solve them. OER creators lack time, skills, and awareness to create accessible OERs. Support from specialists and colleagues and hands-on trainings can help cope with these challenges.

Landscape of IoT Securityis a survey paper that overviews the IoT security landscape intending to emphasize the demand for secured IoT-related products and applications. Therefore, this paper derives (a) a list of key challenges of securing IoT devices is determined by examining their particular characteristics, (b) major security objectives for secured IoT systems are defined, (c) a threat taxonomy, which outlines potential security gaps prevalent in current IoT systems, and (d) key countermeasures against the aforementioned threats, for selected IoT security-related technologies available on the market. Considering the vast number of Internet of Things (IoT) devices currently being deployed and the number of devices that are estimated to exist beyond 2025, it is indisputable that IoT will significantly impact society and the economy. IoT devices rapidly become part of everyday life, and all relevant sectors will encounter them. Selected industries are already fully involved in developing IoT applications: factories, medical institutions, homes, and cities. Due to this growth and entanglement into everyday life, IoT failures and attacks can be severe. Hence, IoT security is a concern of extreme significance. Several studies infer that several challenges in securing IoT devices and networks exist. Due to the particular characteristics of IoT devices, it is not feasible to apply traditional IT countermeasures since only dedicated IoT security procedures have to be developed. However, since there is already considerable progress made in relevant fields, especially IoT device hardware, network management, authentication, privacy policies, forensics, and lifecycle management, the path to secure IoT has been opened. The fast progress of IoT security research identified in this can be supported by various products emerging on the market. While products can be distinguished as Software, Hardware/ Firmware, Service/Cloud, and Home solutions depending on where security measures are applied, for software solutions, the most common approach still refers to Intrusion Detection Systems. Hardware and firmware solutions secure IoT devices without installing additional software and are primarily intended for IoT manufacturers. Service and cloud products focus on securing the entire network of distributed IoT devices by checking their configuration and monitoring their behavior to detect unusual actions. In private households, the key objective of IoT security is to protect the users' privacy. Home Solutions can scan data traffic and ensure no sensitive information is leaked. Within the same context, the need for regulations arose. Multiple projects and working groups, such as the Internet Engineering Task Force (IETF), Global System for Mobile Communications Association (GSMA), Open Web Application Security Project (OWASP), or Broadband Internet Technical Advisory Group (BITAG), elaborate best practices for the design, development, and deployment of secure IoT services and products. This shared knowledge makes it possible for smaller manufacturers with negligible expertise and limited resources to offer secure IoT ecosystems. Furthermore, governmental agencies, like the American Homeland Security, the National Institute of Standards and Technology (NIST), or the European Union Agency for Cybersecurity (ENISA), work on regulations or guidelines to protect the population. However, several gaps between existing regulations like General Data Protection Regulation (GDPR) and the IoT still need to be addressed. The threat taxonomy, developed in this paper and based on the literature reviews, provides a more exhaustive view of possible IoT threats and IoT attack vectors than existing taxonomies do. However, while the taxonomy is based on the standard three-layer IoT architecture, the categorization based on layers helps to visualize where a threat can occur, which is limited due to its simplicity. For selected threats, it is not trivial where they should be positioned within the taxonomy. For example, a Distributed Denial of Service (DDoS) threat can affect the network and the application layer. Additionally, the taxonomy does not address that selected threats differing in nature depending on the context: a node within the system can be on the receiving end of a DDoS attack, i.e., if other nodes are sending requests to it, or on the sending end, i.e., if it was compromised and sent requests to other nodes. Additionally, the survey highlights the importance of considering the complete lifecycle of IoT devices when addressing security concerns. However, the taxonomy presented needs to account for the dynamic nature of IoT systems, where devices might join a network at any time and belong to multiple owners during their lifecycle. Thus, such a false impression of addressing IoT security as a single task has to be countermeasured since securing the system needs to be a permanent task. Future versions of the taxonomy developed can address these aspects by incorporating a dynamic view of IoT security. Additionally, they can include a ranking of threats based on metrics, such as the Common Vulnerability Scoring System (CVSS) by NIST. By adding data from public data sets, such as the Common Vulnerabilities and Exposure (CVE) database, which lists known vulnerabilities, an additional perspective can be determined on how common specific threats are. Exploring how these threats can be combined to form a standard attack surface that a malicious user might take will be interesting.

Answering connectivity queries is fundamental to fully dynamic graphs where edges and vertices are inserted and deleted frequently. Existing work proposes data structures and algorithms with worst case guarantees. We propose a new data structure, the dynamic tree (D-tree), together with algorithms to construct and maintain it. The D-tree is the first data structure that scales to fully dynamic graphs with millions of vertices and edges and, on average, answers connectivity queries much faster than data structures with worst case guarantees.

Designing reward functions for reinforcement learning is difficult: besides specifying which behavior is rewarded for a task, the reward also has to discourage undesired outcomes. Misspecified reward functions can lead to unintended negative side effects, and overall unsafe behavior. To overcome this problem, recent work proposed to augment the specified reward function with an impact regularizer that discourages behavior that has a big impact on the environment. Although initial results with impact regularizers seem promising in mitigating some types of side effects, important challenges remain. In this paper, we examine the main current challenges of impact regularizers and relate them to fundamental design decisions. We discuss in detail which challenges recent approaches address and which remain unsolved. Finally, we explore promising directions to overcome the unsolved challenges in preventing negative side effects with impact regularizers., Comment: Presented at the SafeAI workshop at AAAI 2021

Rapidly emerging augmented reality technologies enable us to virtually alter appearance of objects and materials in a fast and efficient way. The state-of-the-art research shows that the human visual system has a poor ability to invert the optical processes in the scene and rather relies on images cues and spatial distribution of luminance to perceive appearance attributes, such as gloss and translucency. For this reason, we hypothesize that it is possible to alter gloss and translucency appearance by projecting an image onto the original to mimic the luminance distribution characteristic of the target appearance. To demonstrate feasibility of this approach, we use pairs of physically-based renderings of glossy and matte, and translucent and opaque materials, respectively; we calculate a compensation image – a luminance difference between them, and subsequently, we demonstrate that by algebraic addition of luminance, an image of matte object can appear glossy, and an image of opaque object can appear translucent, when respective compensation images are projected onto them. Furthermore, we introduce a novel method to increase apparent opacity of translucent materials. Finally, we propose a future direction, which could enable nearly real-time appearance manipulation.

In this paper, we support the relevance of the collaboration and mutual inspiration between research in Artificial Intelligence and neuroscience to create truly intelligent and efficient systems. In contrast to the traditional top-down and bottom-up strategies designed to study and emulate the brain, we propose an alternative approach where these two strategies are met halfway, defining a set of algorithmic principles. We present prediction as a core algorithmic principle and advocate for applying the same approach to identify other neural principles which can constitute core mechanisms of new Machine Learning frameworks., CEUR Workshop Proceedings, 3227, ISSN:1613-0073, Proceedings of the 3rd Human-Like Computing Workshop (HLC 2022)

Previous research investigating the use of technology in school has focused mainly on the frequency of use of digital tools during lessons rather than investigating how technology is integrated with respect to different kinds of learning activities. Since the impact of technology use on learning depends on how it is used and on what activities supported by technology are implemented in lessons, a measurement instrument assessing how technology is integrated into learning activities is necessary to investigate its impact on teaching and learning processes. According to the interactive, constructive, active, and passive (ICAP) framework, which distinguishes four different learning activities based on the level of students’ cognitive engagement, we developed the 12-item ICAP Technology Scale (ICAP-TS) that accounts for all four dimensions of technology integration in lessons. We used confirmatory factor analysis to validate the four-factor structure of the ICAP-TS with a sample of 1059 upper-secondary school teachers from Switzerland. We also examined reliability using classical test theory and Rasch model analysis to assess the scale’s psychometric characteristics. We then analyzed the associations between the ICAP-TS and a general use frequency measure of 12 educational technologies to test the criterion validity. The results confirmed the four-factor structure of the ICAP-TS and revealed good instrument accuracy. The most difficult items to endorse are those describing the integration of technology into interactive learning activities. Furthermore, all 12 items significantly correlated with the frequency of use of 12 educational technologies. We recommend the ICAP-TS as a short and reliable measurement scale for assessing how technology is integrated into lessons, considering different learning activities based on the ICAP theoretical model.

Sickle cell disease (SCD) is a severe genetic hemoglobin disorder that results in premature destruction of red blood cells. Assessment of the severity of the disease is a challenging task in clinical routine since the causes of broad variance in SCD manifestation despite the common genetic cause remain unclear. Identification of the biomarkers that would predict the severity grade is of importance for prognosis and assessment of responsiveness of patients to therapy. Detection of the changes in red blood cell (RBC) density through separation of Percoll density gradient could be such marker as it allows to resolve intercellular differences and follow the most damaged dense cells prone to destruction and vaso-occlusion. Quantification of the images obtained from the distribution of RBCs in Percoll gradient and interpretation of the obtained is an important prerequisite for establishment of this approach. Here, we propose a novel approach combining a graph convolutional network, a convolutional neural network, fast Fourier transform, and recursive feature elimination to predict the severity of SCD directly from a Percoll image. Two important but expensive laboratory blood test parameters measurements are used for training the graph convolutional network. To make the model independent from such tests during prediction, the two parameters are estimated by a neural network from the Percoll image directly. On a cohort of 216 subjects, we achieve a prediction performance that is only slightly below an approach where the groundtruth laboratory measurements are used. Our proposed method is the first computational approach for the difficult task of SCD severity prediction. The two-step approach relies solely on inexpensive and simple blood analysis tools and can have a significant impact on the patients' survival in underdeveloped countries where access to medical instruments and doctors is limited, Accepted for publication at MICCAI 2021 workshop on aFfordable healthcare and AI for Resource diverse global health (FAIR)

Datasets and methods for cross-document coreference resolution (CDCR) focus on events or entities with strict coreference relations. They lack, however, annotating and resolving coreference mentions with more abstract or loose relations that may occur when news articles report about controversial and polarized events. Bridging and loose coreference relations trigger associations that may lead to exposing news readers to bias by word choice and labeling. For example, coreferential mentions of "direct talks between U.S. President Donald Trump and Kim" such as "an extraordinary meeting following months of heated rhetoric" or "great chance to solve a world problem" form a more positive perception of this event. A step towards bringing awareness of bias by word choice and labeling is the reliable resolution of coreferences with high lexical diversity. We propose an unsupervised method named XCoref, which is a CDCR method that capably resolves not only previously prevalent entities, such as persons, e.g., "Donald Trump," but also abstractly defined concepts, such as groups of persons, "caravan of immigrants," events and actions, e.g., "marching to the U.S. border." In an extensive evaluation, we compare the proposed XCoref to a state-of-the-art CDCR method and a previous method TCA that resolves such complex coreference relations and find that XCoref outperforms these methods. Outperforming an established CDCR model shows that the new CDCR models need to be evaluated on semantically complex mentions with more loose coreference relations to indicate their applicability of models to resolve mentions in the "wild" of political news articles.

We are interested in building schemes for the compressible Euler equations that are also locally conserving the angular momentum. We present a general framework, describe a few examples of schemes and show results. These schemes can be of arbitrary order.

Information propagation in online social networks is facilitated by two types of influence - endogenous (peer) influence that acts between users of the social network and exogenous (external) that corresponds to various external mediators such as online news media. However, inference of these influences from data remains a challenge, especially when data on the activation of users is scarce. In this paper we propose a methodology that yields estimates of both endogenous and exogenous influence using only a social network structure and a single activation cascade. Our method exploits the statistical differences between the two types of influence - endogenous is dependent on the social network structure and current state of each user while exogenous is independent of these. We evaluate our methodology on simulated activation cascades as well as on cascades obtained from several large Facebook political survey applications. We show that our methodology is able to provide estimates of endogenous and exogenous influence in online social networks, characterize activation of each individual user as being endogenously or exogenously driven, and identify most influential groups of users.

Internet usage has grown dramatically since the early years of its inception. The rich field of data provided by internet users in interaction with digital media content can provide insight into web-based navigation behavior and underlying psychological dimensions. Human-computer interaction in the web is an underutilized source of data for understanding human online behavior. While researchers and usability testing services do use these sources to analyze human behavior and user experience, access to the diverse range of other potentially useful data available during web-based interaction for research is limited. In this paper, we propose a novel tool in the form of a web browser extension, referred to as Latent, which can be used to simultaneously capture information from different sources while users interact with digital content. The data acquisition capabilities of Latent makes it suitable for various research purposes, ranging from studies of usability to decision-making and personality. A particular advantage of Latent is that the method and control of data acquisition is completely transparent to the user. We present the architecture of the web browser extension, describe the data that can be acquired, and report on the residual impact of the tool on the user's computer processing resources.

Radiomic representations can quantify properties of regions of interest in medical image data. Classically, they account for pre-defined statistics of shape, texture, and other low-level image features. Alternatively, deep learning-based representations are derived from supervised learning but require expensive annotations from experts and often suffer from overfitting and data imbalance issues. In this work, we address the challenge of learning representations of 3D medical images for an effective quantification under data imbalance. We propose a \emph{self-supervised} representation learning framework to learn high-level features of 3D volumes as a complement to existing radiomics features. Specifically, we demonstrate how to learn image representations in a self-supervised fashion using a 3D Siamese network. More importantly, we deal with data imbalance by exploiting two unsupervised strategies: a) sample re-weighting, and b) balancing the composition of training batches. When combining our learned self-supervised feature with traditional radiomics, we show significant improvement in brain tumor classification and lung cancer staging tasks covering MRI and CT imaging modalities., camera-ready version in MICCAI 2021

In this work, we propose a method utilizing tool-integrated vibroacoustic measurements and a spatio-temporal learning-based framework for the detection of the insertion endpoint during femoral stem implantation in cementless Total Hip Arthroplasty (THA). In current practice, the optimal insertion endpoint is intraoperatively identified based on surgical experience and dependent on a subjective decision. Leveraging spectogram features and time-variant sequences of acoustic hammer blow events, our proposed solution can give real-time feedback to the surgeon during the insertion procedure and prevent adverse events in clinical practice. To validate our method on real data, we built a realistic experimental human cadaveric setup and acquired acoustic signals of hammer blows during broaching the femoral stem cavity with a novel inserter tool which was enhanced by contact microphones. The optimal insertion endpoint was determined by a standardized preoperative plan following clinical guidelines and executed by a board-certified surgeon. We train and evaluate a Long-Term Recurrent Convolutional Neural Network (LRCN) on sequences of spectrograms to detect a reached target press fit corresponding to a seated implant. The proposed method achieves an overall per-class recall of \(93.82\pm 5.11\%\) for detecting an ongoing insertion and \(70.88\pm 11.83\%\) for identifying a reached target press fit for five independent test specimens. The obtained results open the path for the development of automated systems for intra-operative decision support, error prevention and robotic applications in hip surgery.

This is a part of an in-depth study of a set of the manuscripts related to the Jayadrathayāmala. Taking JY.3.9 as a test-chapter, a comparative paleography analysis of the 11 manuscripts was made within READ software framework. The workflow within READ minimized the effort to make a few important discoveries (manuscripts containing more than one script, identification of the manuscripts potentially written by the same person) as well as to create an overview of the shift from Nāgarī to Newārī and, finally, to Devanāgarī scripts within the history of manuscript transmission of a single chapter. Exploratory statistical analysis in R of the syllable frequency in each manuscript, based on the paleography analysis export from READ, helped to establish that there are potentially two lines of manuscript transmission of the JY.3.9.

In this short paper, we discuss a straight-forward approach for the identification of noun phrases denoting actors (agents). We use a multilayer perceptron applied to the word embeddings of the head nouns in order to learn a model. A list of 9,000 actors together with 11,000 non-actors generated from a newspaper corpus are used as a silver standard. An evaluation of the results seems to indicate that the model generalises well on unseen data.

Learning the cumulative distribution function (CDF) of an outcome variable conditional on a set of features remains challenging, especially in high-dimensional settings. Conditional transformation models provide a semi-parametric approach that allows to model a large class of conditional CDFs without an explicit parametric distribution assumption and with only a few parameters. Existing estimation approaches within this class are, however, either limited in their complexity and applicability to unstructured data sources such as images or text, lack interpretability, or are restricted to certain types of outcomes. We close this gap by introducing the class of deep conditional transformation models which unifies existing approaches and allows to learn both interpretable (non-)linear model terms and more complex neural network predictors in one holistic framework. To this end we propose a novel network architecture, provide details on different model definitions and derive suitable constraints as well as network regularization terms. We demonstrate the efficacy of our approach through numerical experiments and applications.

Virtual Reality (VR) has emerged and developed as a new modality to interact with multimedia data. In this paper, we present vitrivr-VR, a prototype of an interactive multimedia retrieval system in VR based on the open source full-stack multimedia retrieval system vitrivr. We have implemented query formulation tailored to VR: Users can use speech-to-text to search collections via text for concepts, OCR and ASR data as well as entire scene descriptions through a video-text co-embedding feature that embeds sentences and video sequences into the same feature space. Result presentation and relevance feedback in vitrivr-VR leverages the capabilities of virtual spaces.

Autonomous driving solutions are based on artificial vision and machine learning for understanding the environment and facilitate decision making tasks. Similar techniques are used for indoor robot navigation. Deep learning architectures, which are usually computationally expensive, are impacting our daily lives. This technology is evolving with a notable improvement of cost-efficiency in terms of energy consumption, enabling AI-edge computing. However, these architectures are usually trained on powerful GPUs, what represents the limit for edge computing. Nevertheless, after this training, efficient edge computing devices can process these architectures locally. Neuromorphic engineering shows off on solving the energy bottleneck problem through bio-inspired sensors, processors and spike-based computation techniques. This work presents a mobile robotic platform commanded through the Robotic Operating System (ROS), which obeys the classification output of an AI-edge CNN accelerator for FPGA connected to a neuromorphic dynamic vision sensor. The classification system is able to process up to 200 fps for 64 \(\times \) 64 histograms collected with 2k events per frame and executing a 5 layer CNN with 18MOPs for indoor robot navigation. A traffic sign dataset has been used for training achieving a measured accuracy of 97.62% and 99.96% in the validation and test datasets respectively.

Automated detection of mitotic figures in histopathology images is a challenging task: here, we present the different steps that describe the strategy we applied to participate in the MIDOG 2021 competition. The purpose of the competition was to evaluate the generalization of solutions to images acquired with unseen target scanners (hidden for the participants) under the constraint of using training data from a limited set of four independent source scanners. Given this goal and constraints, we joined the challenge by proposing a straight-forward solution based on a combination of state-of-the-art deep learning methods with the aim of yielding robustness to possible scanner-related distributional shifts at inference time. Our solution combines methods that were previously shown to be efficient for mitosis detection: hard negative mining, extensive data augmentation, rotation-invariant convolutional networks. We trained five models with different splits of the provided dataset. The subsequent classifiers produced F1-scores with a mean and standard deviation of 0.747+/-0.032 on the test splits. The resulting ensemble constitutes our candidate algorithm: its automated evaluation on the preliminary test set of the challenge returned a F1-score of 0.6828.

Web search engines influence perception of social reality by filtering and ranking information. However, their outputs are often subjected to bias that can lead to skewed representation of subjects such as professional occupations or gender. In our paper, we use a mixed-method approach to investigate presence of race and gender bias in representation of artificial intelligence (AI) in image search results coming from six different search engines. Our findings show that search engines prioritize anthropomorphic images of AI that portray it as white, whereas non-white images of AI are present only in non-Western search engines. By contrast, gender representation of AI is more diverse and less skewed towards a specific gender that can be attributed to higher awareness about gender bias in search outputs. Our observations indicate both the the need and the possibility for addressing bias in representation of societally relevant subjects, such as technological innovation, and emphasize the importance of designing new approaches for detecting bias in information retrieval systems., Comment: 16 pages, 3 figures

Reminiscence-the act of recalling or telling others about relevant personal past experiences-plays an important role in the well-being of older adults. Therefore, it is relevant to develop intelligent systems aiming at improving the well-being of the elderly by reliably detecting reminiscence in their everyday life conversations. Data imbalance is one of the main challenges in the automatic detection of reminiscence from everyday conversations, as reminiscing is a rare event. In this paper, we address the problem by proposing a methodology for coping with imbalanced data in the detection of reminiscence in conversations of older adults. The methodology combines data augmentation using BERT (Bidirectional Encoder Representations from Transformer) and feature extraction techniques leveraging natural language processing for the German language. We evaluate the proposed methodology on a dataset comprising transcripts of social conversations of older adults held in German. We compare our results with a previous work addressing the problem on the same dataset and we show that our approach strongly outperforms the baseline. The results in this study may support the development of intelligent systems for the real-time detection of reminiscence in everyday life of older adults and the design of digital health interventions to support their well-being., IEEE Access, 9, ISSN:2169-3536

Video is a very expressive medium, able to capture a wide variety of information in different ways. While there have been many advances in the recent past, which enable the annotation of semantic concepts as well as individual objects within video, their larger context has so far not extensively been used for the purpose of retrieval. In this paper, we introduce the first iteration of VideoGraph, a knowledge graph-based video retrieval system. VideoGraph combines information extracted from multiple video modalities with external knowledge bases to produce a semantically enriched representation of the content in a video collection, which can then be retrieved using graph traversal. For the 2021 Video Browser Showdown, we show the first proof-of-concept of such a graph-based video retrieval approach.

The evaluation of the performance of interactive multimedia retrieval systems is a methodologically non-trivial endeavour and requires specialized infrastructure. Current evaluation campaigns have so far relied on a local setting, where all retrieval systems needed to be evaluated at the same physical location at the same time. This constraint does not only complicate the organization and coordination but also limits the number of systems which can reasonably be evaluated within a set time frame. Travel restrictions might further limit the possibility for such evaluations. To address these problems, evaluations need to be conducted in a (geographically) distributed setting, which was so far not possible due to the lack of supporting infrastructure. In this paper, we present the Distributed Retrieval Evaluation Server (DRES), an open-source evaluation system to facilitate evaluation campaigns for interactive multimedia retrieval systems in both traditional on-site as well as fully distributed settings which has already proven effective in a competitive evaluation.

Extensive research on target-dependent sentiment classification (TSC) has led to strong classification performances in domains where authors tend to explicitly express sentiment about specific entities or topics, such as in reviews or on social media. We investigate TSC in news articles, a much less researched domain, despite the importance of news as an essential information source in individual and societal decision making. This article introduces NewsTSC, a manually annotated dataset to explore TSC on news articles. Investigating characteristics of sentiment in news and contrasting them to popular TSC domains, we find that sentiment in the news is expressed less explicitly, is more dependent on context and readership, and requires a greater degree of interpretation. In an extensive evaluation, we find that the state of the art in TSC performs worse on news articles than on other domains (average recall AvgRec = 69.8 on NewsTSC compared to AvgRev = [75.6, 82.2] on established TSC datasets). Reasons include incorrectly resolved relation of target and sentiment-bearing phrases and off-context dependence. As a major improvement over previous news TSC, we find that BERT's natural language understanding capabilities capture the less explicit sentiment used in news articles.

Many researchers have investigated how barriers to technology integration affect the use of digital technologies in teaching and learning. However, the results have varied across educational contexts and countries. Large-scale assessment studies have described barriers only on a descriptive level instead of analyzing the effects of barriers on actual indicators of technology integration, such as technology use. Therefore, this study investigated the effects of barriers on technology use through the lens of the “will, skill, tool” model (WST model) in different European countries while taking the countries’ technological development level into account. A regression analysis showed that barriers had only a minor impact on the frequency of technology use in the classroom in the large majority of countries. In accordance with theoretical expectations, we found country-specific patterns, with a higher negative impact of technological barriers in less technologically developed countries and teacher-belief related barriers prevalent in developed countries. These findings may help policy makers identify needed interventions in different contexts.