Abstract
Sensors are susceptible to failure when exposed to extreme conditions over long periods of time. Besides they can be affected by noise or electrical interference. Models (Machine Learning or others) obtained from these faulty and noisy sensors may be less reliable. In this paper, we propose a data augmentation approach for making neural networks more robust to missing and faulty sensor data. This approach is shown to be effective in a real life industrial application that uses data of various sensors to predict the wear of an automotive fuel-system component. Empirical results show that the proposed approach leads to more robust neural network in this particular application than existing methods.

Abstract
This paper presents an optimal bidding and offering strategy for a virtual power plant (VPP), which participates in day-ahead (DA) and balancing markets. The VPP comprises distributed energy resources, plug-in electric vehicles (PEVs) and flexible demands. The objective of the problem is maximizing the VPP's profit while demand response (DR) providers who aggregated the loads try to supply the required demand under their jurisdiction with minimum costs. The proposed optimization problem is formulated as a bi-level stochastic scheduling programming to address uncertainties in DA and balancing electricity prices, renewable energy source's (RES) and DR relationship. Simulation results demonstrate the applicability and effectiveness of the proposed model to any real markets. Also, numerical results show that the flexibility of responsive loads and PEVs can improve the VPP operator's energy management and increase its expected profit. © 2020 IEEE.

Abstract
Objectives: To develop and validate a clinical model to identify patients admitted to hospital with community-acquired infection (CAI) caused by pathogens resistant to antimicrobials recommended in current CAI treatment guidelines. Methods: International prospective cohort study of consecutive patients admitted with bacterial infection. Logistic regression was used to associate risk factors with infection by a resistant organism. The final model was validated in an independent cohort. Results: There were 527 patients in the derivation and 89 in the validation cohort. Independent risk factors identified were: atherosclerosis with functional impairment (Karnofsky index <70) [adjusted OR (aOR) (95% CI) = 2.19 (1.41-3.40)]; previous invasive procedures [adjusted OR (95% CI) = 1.98 (1.28-3.05)]; previous colonization with an MDR organism (MDRO) [aOR (95% CI) = 2.67 (1.48-4.81)]; and previous antimicrobial therapy [aOR (95% CI) = 2.81 (1.81-4.38)]. The area under the receiver operating characteristics (AU-ROC) curve (95% CI) for the final model was 0.75 (0.70-0.79). For a predicted probability >= 22% the sensitivity of the model was 82%, with a negative predictive value of 85%. In the validation cohort the sensitivity of the model was 96%. Using this model, unnecessary broad-spectrum therapy would be recommended in 30% of cases whereas undertreatment would occur in only 6% of cases. Conclusions: For patients hospitalized with CAI and none of the following risk factors: atherosclerosis with functional impairment; previous invasive procedures; antimicrobial therapy; or MDRO colonization, CAI guidelines can safely be applied. Whereas, for those with some of these risk factors, particularly if more than one, alternative antimicrobial regimens should be considered.

Abstract
Time series of (small) counts are common in practice and appear in a wide variety of fields. In the last three decades, several models that explicitly account for the discreteness of the data have been proposed in the literature. However, for multivariate time series of counts several difficulties arise and the literature is not so detailed. This work considers Bivariate INteger-valued Moving Average, BINMA, models based on the binomial thinning operation. The main probabilistic and statistical properties of BINMA models are studied. Two parametric cases are analysed, one with the cross-correlation generated through a Bivariate Poisson innovation process and another with a Bivariate Negative Binomial innovation process. Moreover, parameter estimation is carried out by the Generalized Method of Moments. The performance of the model is illustrated with synthetic data as well as with real datasets.

Abstract
There is an ongoing paradigm shift occurring in the electricity sector. In particular, previously passive consumers are now becoming active prosumers and they can now offer important and cost-effective new forms of flexibility and demand response potential to the electricity sector and this can translate into system-wide operational and economic benefits. This work focuses on developing a model where prosumers participate in demand response programs through varying tariff schemes, and the model also quantifies the benefits of this flexibility and cost-reductions. This work includes transactive energy trading between various prosumers, the grid and the neighborhood. A stochastic tool is developed for this analysis, which also allows the quantification of the collective behavior so that the periods with the greatest demand response potential can be identified. Numerical results indicate that the optimization of energy transactions amongst the prosumers, and including the grid, leads to considerable cost reductions as well as introducing additional flexibility in the presence of demand response mechanisms.

Abstract
Fingerprint presentation attack detection (PAD) methods present a stunning performance in current literature. However, the fingerprint PAD generalisation problem is still an open challenge requiring the development of methods able to cope with sophisticated and unseen attacks as our eventual intruders become more capable. This work addresses this problem by applying a regularisation technique based on an adversarial training and representation learning specifically designed to to improve the PAD generalisation capacity of the model to an unseen attack. In the adopted approach, the model jointly learns the representation and the classifier from the data, while explicitly imposing invariance in the high-level representations regarding the type of attacks for a robust PAD. The application of the adversarial training methodology is evaluated in two different scenarios: i) a handcrafted feature extraction method combined with a Multilayer Perceptron (MLP); and ii) an end-to-end solution using a Convolutional Neural Network (CNN). The experimental results demonstrated that the adopted regularisation strategies equipped the neural networks with increased PAD robustness. The adversarial approach particularly improved the CNN models' capacity for attacks detection in the unseen-attack scenario, showing remarkable improved APCER error rates when compared to state-of-the-art methods in similar conditions.