Progress in making neural networks more robust against adversarial attacks is mostly marginal, despite the great efforts of the research community. Moreover, the robustness evaluation is often imprecise, making it challenging to identify promising approaches. We do an observational study on the classification decisions of 19 different state-of-the-art neural networks trained to be robust against adversarial attacks. This analysis gives a new indication of the limits of the robustness of current models on a common benchmark. In addition, our findings suggest that current untargeted adversarial attacks induce misclassification toward only a limited amount of different classes. Similarly, we find that previous attacks under-explore the perturbation space during optimization. This leads to unsuccessful attacks for samples where the initial gradient direction is not a good approximation of the final adversarial perturbation direction. Additionally, we observe that both over- and under-confidence in model predictions result in an inaccurate assessment of model robustness. Based on these observations, we propose a novel loss function for adversarial attacks that consistently improves their efficiency and success rate compared to prior attacks for all 30 analyzed models.

Ho Chi Minh City (HCMC, Vietnam) is one of the fastest growing megacities in the world. In this paper, we attempt to analyse the dynamics of nutrients, suspended sediments, and water discharges in its aquatic systems today and in the future. The work is based on nine sampling sites along the Saigon River and one on the Dongnai River to identify the reference water status upstream from the urban area and the increase in fluxes that occur within the city and its surroundings. For the first time, the calculated fluxes allow drawing up sediment and nutrient budgets at the basin scale and the quantification of total nutrient loading to the estuarine and coastal zones (2012–2016 period). Based on both national Vietnamese and supplementary monitoring programs, we estimated the water, total suspended sediment, and nutrients (Total N, Total P, and dissolved silica: DSi) fluxes at 137 m³ year⁻¹, 3,292 × 10³ tonSS year⁻¹, 5,323 tonN year⁻¹, 450 tonP year⁻¹, and 2,734 tonSi year⁻¹ for the Saigon River and 1,693 m³ year⁻¹, 1,175 × 10³ tonSS year⁻¹, 31,030 tonN year⁻¹, 1,653 tonP year⁻¹, and 31,138 tonSi year⁻¹ for the Dongnai River, respectively. Nutrient fluxes provide an indicator of coastal eutrophication potential (indicator of coastal eutrophication potential), using nutrient stoichiometry ratios. Despite an excess of nitrogen and phosphorus over silica, estuarine waters downstream of the megacity are not heavily impacted by HCMC. Finally, we analysed scenarios of future trends (2025–2050) for the nutrient inputs on the basis of expected population growth in HCMC and improvement of wastewater treatment capacity. We observed that without the construction of a large number of additional wastewater treatment plants, the eutrophication problem is likely to worsen. The results are discussed in the context of the wastewater management policy.

Gait symptoms and balance impairment are characteristic indicators for the progression in Parkinson’s disease (PD). Current gait assessments mostly focus on straight strides with assumed constant velocity, while acceleration/deceleration and turning strides are often ignored. This is either due to the set up of typical clinical assessments or technical limitations in capture volume. Wearable inertial measurement units are a promising and unobtrusive technology to overcome these limitations. Other gait phases such as initiation, termination, transitioning (between straight walking and turning) and turning might be relevant as well for the evaluation of gait and balance impairments in PD. In a cohort of 119 PD patients, we applied unsupervised algorithms to find different gait clusters which potentially include the clinically relevant information from distinct gait phases in the standardized 4x10 m gait test. To clinically validate our approach, we determined the discriminative power in each gait cluster to classify between impaired and unimpaired PD patients and compared it to baseline (analyzing all straight strides). As a main result, analyzing only one of the gait clusters constant, non-constant or turning led in each case to a better classification performance in comparison to the baseline (increase of area under the curve (AUC) up to 19% relative to baseline). Furthermore, gait parameters (for turning, constant and non-constant gait) that best predict motor impairment in PD were identified. We conclude that a more detailed analysis in terms of different gait clusters of standardized gait tests such as the 4x10 m walk may give more insights about the clinically relevant motor impairment in PD patients.

The most important goal of customer service is to keep the customer satisfied. However, service resources are always limited and must prioritize specific customers. Therefore, it is essential to identify customers who potentially become unsatisfied and might lead to escalations . Data science on IoT data (especially log data) for machine health monitoring and analytics on enterprise data for customer relationship management (CRM) have mainly been researched and applied independently. This paper presents a data-driven decision support system framework that combines IoT and enterprise data to model customer sentiment and predicts escalations. The proposed framework includes a fully automated and interpretable machine learning pipeline using state-of-the-art methods. The framework is applied in a real-world case study with a major medical device manufacturer providing data from a fleet of thousands of high-end medical devices. An anonymized version of this industrial benchmark is released for the research community based on the presented case study, which has interesting and challenging properties. In our extensive experiments, we achieve a Recall@50 of 50.0 % for the task of predicting customer escalations. In addition, we show that combining IoT and enterprise data can improve prediction results and ease troubleshooting. Additionally, we propose a practical workflow for end-users when applying the proposed framework.