Abstract
In a time in which streaming data becomes the new normal in Machine Learning problems, to the detriment of batch data, new challenges arise. In the past, a data source would be static in the sense that all data were known at the moment of the training of the model. A model would be trained and it would be in use for relatively long periods of time. Nowadays, data arrive in real-time and their statistical properties may also change over time, rendering trained models outdated. In this paper we propose an approach to deal with the concept drift problem with minimal computational effort. Specifically, we continuously update an ensemble with new weak learners and adjust their weights according to their performance. This approach is suitable to be used in real-time in the form of an ever-evolving model that adapts to change in the data.

Abstract
Federated Learning (FL) is a distributed learning methodology that allows multiple nodes to cooperatively train a deep learning model, without the need to share their local data. It is a promising solution for telemonitoring systems that demand intensive data collection, for detection, classification, and prediction of future events, from different locations while maintaining a strict privacy constraint. Due to privacy concerns and critical communication bottlenecks, it can become impractical to send the FL updated models to a centralized server. Thus, this paper studies the potential of hierarchical FL in Internet of Things (IoT) heterogeneous systems. In particular, we propose an optimized solution for user assignment and resource allocation over hierarchical FL architecture for IoT heterogeneous systems. This work focuses on a generic class of machine learning models that are trained using gradient-descent-based schemes while considering the practical constraints of non-uniformly distributed data across different users. We evaluate the proposed system using two real-world datasets, and we show that it outperforms state-of-the-art FL solutions. Specifically, our numerical results highlight the effectiveness of our approach and its ability to provide 4–6% increase in the classification accuracy, with respect to hierarchical FL schemes that consider distance-based user assignment. Furthermore, the proposed approach could significantly accelerate FL training and reduce communication overhead by providing 75–85% reduction in the communication rounds between edge nodes and the centralized server, for the same model accuracy.

Abstract
This work aims to systematize the knowledge on emerging Intelligent Information Retrieval (IIR) practices in scenarios whose context is similar to the field of tax law. It is a part of a project that covers the emerging techniques of IIR and its applicability to the tax law domain. Furthermore, it presents an overview of different approaches for representing legal data and exposes the challenging task of providing quality insights to support decision-making in a dedicated legal environment. It also offers an overview of the related background and prior research referring to the techniques for information retrieval in legal documents, establishing the current state-of-the-art, and identifying its main drawbacks. A summary of the most appropriate technologies and research approaches of the technologies that apply artificial intelligence technology to help legal tasks is also depicted.

Abstract
Convolutional Neural Networks (CNN) have been successfully applied to images, text and audio, but their performance are not so good when applied to feature-based tabular data. Exceptions are works such as TabNet and DeepInsight, which employ end-to-end approaches. In this work, we propose an alternative way of using CNNs to model tabular data where knowledge is extracted from the feature space before being introduced to the network. Our strategy, Map-Optimize-Learn (MOL), changes the shape representation of samples in order to produce suitable input data for the CNN architecture. The strategy is applied to a real-world scenario of children and teenagers with cardiac pathology and compared against baseline and state of the art Machine Learning (ML) algorithms for tabular datasets. Preliminary results suggest that the strategy has potential to improve prediction quality of tabular data over end-to-end CNN methods and classical ML methods.

Abstract
Wastewater treatment systems are major consumers of electricity being responsible for 3 to 5% of global energy consumption, and 56% of greenhouse gas emissions into the atmosphere in the water treatment sector. Climate change currently imposes the definition of a new pattern of human behavior in the defense and sharing of a common space that is the planet, so the optimization of water treatment models plays a crucial role in the definition of sustainability strategies as part of the challenges for decarbonization by 2050. The physical-chemical characteristics of the influent, the treatment techniques and associated technologies and the unpredictability of external phenomena of inefficiency transform wastewater treatment plants (WWTPs) into complex systems, sometimes difficult to understand. The study of energy efficiency plays an important role in the emergence of a standard behavior model, which allows the correction of unbalanced situations in the expected energy consumption. Given the importance of the topic, the present review aims to study energy auditing techniques and benchmarking tools developed for the wastewater treatment sector to reduce the current electricity consumption, which could represent up to 90% of total energy consumption. The result of the research was organized according to the criteria defined for the characterization of auditing techniques and benchmarking tools. A review was conducted from 51 scientific papers from different reference research platforms published in the last 20 years according to the keywords. This literature review has shown that there are, in the classification of consumption reduction, energy auditing and benchmarking tools; energy management techniques and methods directed to the energy efficiency of the treatment stages and specific equipment; and, finally, decision support tools. According to the methodology followed, it was possible to conclude that although the concern is not recent, there are techniques and tools for assessing energy performance more suitable for the wastewater sector. However, the authors recognize that associated with the complexity of wastewater treatment systems, inefficiency phenomena still strongly impact energy efficiency assessment, so the contributions for their identification and quantification may represent an added value for data analysis, systematization, and optimization methodologies.

Abstract