Audiometer systems provide enormous amounts of detailed TV watching data. Several relevant and interdependent factors may influence TV viewers' behavior. In this work we focus on the time factor and derive Temporal Patterns of TV watching, based on panel data. Clustering base attributes are originated from 1440 binary minute-related attributes, capturing the TV watching status (watch/not watch). Since there are around 2500 panel viewers a data reduction procedure is first performed. K-Means algorithm is used to obtain daily clusters of viewers. Weekly patterns are then derived which rely on daily patterns. The obtained solutions are tested for consistency and stability. Temporal TV watching patterns provide new insights concerning Portuguese TV viewers' behavior.

A personal collection of photos shows large variability in the depicted items, making difficult a fully automated solution to cope with sensory and semantic gaps. Emotions and non-visual contextual information can be very important to address those problems. Manual annotations are key, but their time-consuming nature alienate users from doing them. One solution is to lower the annotation effort, building solutions on top of algorithms that prepare a context separation, making possible the reuse of annotations. In this paper we present a segmentation algorithm that uses spatio-temporal information to segment personal photo collections. The algorithm is assessed in a user study, using the participants own photos. The results show users make none or few changes to the proposed segmentations, indicating an acceptance of the algorithm outcome.

This article is build on the experience of using Google Earth Engine as a development framework for a previous work by the same authors.Being primarily a distributed parallel computing platform, it is designed around a functional language pattern, even though supported on an object model, and a map / reduce distributed workload paradigm.Leveraging the sheer computing power delivered by the Google infrastructure and a multi petabyte remote sensing data repository, GoogleEarth Engine is an efficient development framework that presents itself in two basic flavors: one online integrated development environment which uses the browser Javascript engine; two APIs that can be deployed to a Python or a NodeJS environment.This work emphasizes the comparison between the Javascript browserbased implementation and the Python environment packages.

Web Archiving (WA) deals with the preservation of portions of the World Wide Web (WWW) allowing their availability for future access. Arquivo.pt is a WA initiativeholding a huge amount of content, including image files.However, some of these images contain nudity and pornography, that can be offensive for the users, and thus being Not SuitableFor Work (NSFW). This work proposes a methodology to classify NSFW images available at Arquivo.pt, using deep neural network approaches. A large dataset of images is built using Arquivo.pt data and two pre-trained neural network models, namely ResNet and SqueezeNet, are evaluated and improved for the NSFW classification task, using the dataset.The evaluation of these models reported an accuracy of 93% and 72%, respectively. After a fine tuning stage, the accuracy of these models improved to 94% and 89%, respectively.The proposed solution is integrated into the Arquivo.pt Image Search System, enabling the filtering of the problematic NSFW images. At the time of this writing, the proposed solution is in production at https://arquivo.pt/images.jsp

A poluição atmosférica afeta a saúde da população e é um factor de aumento da taxa de mortalidade. Entidades responsáveis pela gestão de grandes cidades procuram soluções que permitam monitorizar em tempo real os níveis de poluição em grandes áreas urbanas. No entanto, as soluções existentes para a monitorização de poluição são proprietárias ou não apresentam um grau de maturidade de integração, que facilite a expansão de uma rede de sensores com custos baixos de manutenção. Neste trabalho apresenta-se uma arquitectura da solução e uma implementação de referência, para uma plataforma distribuída de código aberto, que permita gerir uma rede de sensores. Esta plataforma permite acoplar e desacoplar, em tempo real, os sensores de poluição atmosférica, armazenar o histórico de medições registadas, agregar dados e disponibilizar um dashboard que permita posterior análise da informação armazenada. A arquitetura de solução é disponibilizada em contentores de execução virtual, recorrendo a diversos componentes do quadro de referência FIWARE e a componentes desenvolvidos à medida. O meio de comunicação usado na transmissão de dados entre os sensores e a plataforma, é uma rede baseada em LoRaWAN, a qual permite comunicações sem fios de longo alcance e de baixa potência. Foi assim selecionada a The Things Network (TTN), uma rede pública que implementa o protocolo LoRaWAN, que permite a comunicação entre sensores e a plataforma proposta. Foi selecionado o protocolo MQTT como forma de comunicação entre a plataforma proposta e a TTN. A plataforma proposta disponibiliza uma interface com o utilizador, implementada como um dashboard acessível a partir de um navegador, onde é possível observar a localização dos sensores registados sobre um mapa, a informação sobre seu estado operacional, bem como a informação sobre as medições observadas. Todos os componentes da plataforma são parametrizados através de um único ficheiro de configuração, que, através de um processo de automação, possibilita a instalação e execução da plataforma através de um único comando.

Finding a new home in a large city has increasingly become more complex, as people are concerned with not only the estate itself but also with its surroundings. The 15-Minute City concept, thinks of cities as a way to ensure that their residents within a 15-minute radius, can fulfill six essential functions: home, work, commerce, health care, education, and entertainment; which varies according to a chosen travel mode. To help people find properties that would fit them according to this concept, we have built an application that aims to provide an intuitive user interface that guides the user through the important decision of buying a house. To present relevant information to the user, we gathered information from three relevant sources, each with its unique challenges. The first source is estate data, extracted from the websites of local real estate agencies; second, city data, with points of interest relevant to the six essential functions mentioned previously; and at last, the user data provided by the user itself through our profile system which lets us understand his needs. System-wise, we built a reliable application following the microservices architecture guidelines, which future proofs our solution by segregating each part of the project and allowing it to scale easily, development and feature wise. The application scrapes, ingests, transforms and stores data regularly. The complete infrastructure is containerized using Docker and orchestrated by Kubernetes. With this application, we provided a scalable solution that allows users to select the best real estate taking into consideration the surrounding environment, tailored to their day-to-day needs, something that, as far as we know, is a novelty on real estate agency websites.

Kolmogorov-Arnold Networks (KANs) represent a breakthrough in deep learning, diverging from Multi-Layer Perceptrons (MLPs) by generalizing the Kolmogorov-Arnold representation theorem (KAT) to networks of arbitrary depth and width. This theorem facilitates the decomposition of multivariate functions into constituent one-dimensional elements, with learnable activation functions on weights and the sum operator on nodes. KANs have been shown to exhibit robust performance in function approximation, validated across mathematical, physical, and practical domains such as traffic prediction and medical diagnostics. Our study investigates KANs’ efficacy through comprehensive evaluations on OpenML, Kaggle and UCI datasets, with a focus on enhancing Human Activity Recognition systems. They demonstrate high classification performance compared to conventional machine learning approaches and MLPs. These findings underscore KANs’ potential as scalable, interpretable tools in modern machine learning applications given their favorable neural scaling laws.

When it comes to web services, RESTful web APIs have become the de facto standard since 2000. Those APIs expose back-end data, so it is crucial that they are robust, secure, and reliable to keep sensitive data protected. Although existing tools for automating APIs test case generation have shown significant potential, they are limited in their applicability since they focus solely on random inputs through fuzzing. Using only API specifications, it is impractical to describe personalized and specific test case workflows. This paper introduces RapiTest, an open-source continuous black-box testing application for RESTful web APIs. It takes advantage of the API specification to automatically generate tests, but also makes use of a new DSL named Test Specification Language (TSL), to create rich test cases. The RapiTest web application allows the setup of several predefined verifications, regarding security and correctness of the responses, while running the tests at regular intervals, such as every 24 hours. In this way, the API can be monitored continuously to ensure it is running correctly.

Web Archiving (WA) deals with the preservation of portions of the World Wide Web (WWW) allowing their availability for future access. Arquivo.pt is a WA initiativeholding a huge amount of content, including image files.However, some of these images contain nudity and pornography, that can be offensive for the users, and thus being Not SuitableFor Work (NSFW). This work proposes a methodology to classify NSFW images available at Arquivo.pt, using deep neural network approaches. A large dataset of images is built using Arquivo.pt data and two pre-trained neural network models, namely ResNet and SqueezeNet, are evaluated and improved for the NSFW classification task, using the dataset.The evaluation of these models reported an accuracy of 93% and 72%, respectively. After a fine tuning stage, the accuracy of these models improved to 94% and 89%, respectively.The proposed solution is integrated into the Arquivo.pt Image Search System, enabling the filtering of the problematic NSFW images. At the time of this writing, the proposed solution is in production at https://arquivo.pt/images.jsp