The objective of this project is to explore the use of visual programming paradigms to enable non-programmers to be part of the Data Science workforce. In contrast to existing approaches, which require programming, Scientific Workflow Management Systems (SWMS) can become an alternative to support the visual programming of data science projects. Such systems (e.g. Taverna and Kepler) use a simple graphical, graph-based structure to develop applications. This simplicity has shown to be suitable in several scientific areas such as bioinformatics, geophysics, and climate analysis. Despite the success of SWMS in data intensive research, they did not reach a state where non-programmers data scientists can use them. They still require some programming and scripting skills to code individual processing tasks. That is why research teams using those systems are usually composed of scientists and software developers. We propose to extend current SWMS to support the parameterization of generic prebuild workflow templates. Workflow templates capture the processing tasks of data science projects. A template can be seen as a formalized best practice that data scientists can use to solve common data analysis challenges. Templates are developed by multidisciplinary teams of experts and reused by non-programmer data scientists. Parameterized workflows have been used successfully in the field of enterprise computing since 1970 to increase software reuse (e.g. SAP’s parameterized workflows to automate business process models). We claim that the same type of benefits can be obtained by parameterizing scientific workflow templates.

A penetração dos dispositivos móveis na sociedade tem levado a que a maioria dos negócios vejam a componente mobile como imprescindível para estar em contacto próximo com os seus clientes. Porém, em 2017, a Google Play Store tinha 2.8 milhões de aplicações móveis disponíveis, a App Store da Apple tinha 2.2 milhões e a Aptoide, copromotora deste projeto, tem atualmente mais de 1 milhão de apps disponíveis, o que gera uma competição extremamente árdua entre apps. Em termos de transações, em 2016 ocorreram 149.3 mil milhões de downloads de apps, número que se espera que duplique em 2020. Porém, muitos destes downloads consistem em várias tentativas para encontrar a aplicação certa, muitas apps transferidas nunca chegam a ser utilizadas e, em 77% dos casos, as apps não voltam a ser utilizadas 72 horas após a sua instalação. Tal demonstra grande desalinhamento entre a oferta de apps por parte das app stores (serviços de distribuição) e a procura das mesmas por parte dos consumidores (descoberta). Devido a este desalinhamento e à muito elevada concorrência entre apps, previsões da Gartner apontam que no fim de 2018 menos de 0.01% dos developers neste mercado considerarão ter atingido sucesso comercial. Além disso, na era cada vez mais digital em que nos encontramos, 52% das apps são descobertas por passa-a-palavra entre conhecidos, amigos ou familiares, e apenas 40% são descobertas pesquisando em app stores. Estas ineficiências fazem da descoberta e distribuição de apps um desafio considerável e extremamente relevante, pois acontecem num mercado de penetração massiva nas sociedades e afetam seriamente a relação entre empresas e consumidores. Partindo deste problema, o projeto AppRecommender tem como objetivo estratégico investigar e desenvolver tecnologias capazes de oferecer a app certa, ao cliente certo, no momento certo, propondo para tal um sistema de recomendações multicritério e um motor de busca semântica. O intuito é otimizar os serviços de distribuição e descobe...

Mobile security faces serious challenges, with alarming threat levels of malicious applications (malware). Malware applications attempt to capture user’s private data for illicit purposes, namely financial data, of personal context (such as location), business / corporate or other kinds of valuable information.  To address this problem the AppSentinel project proposes that App Stores should incorporate proactive and intelligent anti-malware mechanisms themselves, given its privileged position between developers and end-users. In this sense, we propose to research and develop an intelligent anti-malware system for Android App Stores, capable of performing static and dynamic analysis of malicious applications from several sources and understand their behavior patterns, which will then be used in testing new applications submitted to these stores. Moreover, these new applications will also be tested regarding good practices in secure mobile software development, which will lead to educational feedbacks to developers. Finally, a supervised machine learning system will be investigated and developed for efficient detection of new malicious applications based on users’ feedback. With these technological innovations we intend to reduce the incidence of malware on mobile devices, increase the efficiency in the analysis of virus reported by users and accelerate the reaction to new threats, and contribute to the adoption of secure mobile software development practices by developers.

O projeto MESMOQoE assenta no desenvolvimento de uma solução de software (Web Based) direcionada a operadores móveis, capaz de monitorizar, otimizar e prever a utilização de um conjunto de recursos que compõem uma infraestrutura de telecomunicações móveis. Tem como objetivos: 1. Garantir ou melhorar os níveis de Qualidade de Experiência (Quality of Experience, QoE) e Qualidade de Serviço (Quality of Service, QoS) e assumidos pelo operador móvel, através de um processo de auto-optimização dinâmico em malha-fechada.2. Desenvolver modelos e métricas objetivas de avaliação de qualidade, que permitam otimizar a QoE do cliente em vários cenários com especial incidência em aplicações de distribuição multimédia.3. Otimizar a infraestrutura da rede móvel, contribuindo assim para a redução dos custos de operação e aquisição, bem como reduzir o consumo energético da rede.4. Garantir que as necessidades de negócio do operador se encontram alinhadas com o dimensionamento, operação e previsão (forecast) da infraestrutura de rede.

The DeToxSTM aims to develop imaging processing tools able to denoise scanning tunneling microscopy(STM) images. The STM has been considered the most important technique to image,control and monitorize molecular systems with atomic resolution under a wide range of experimentalconditions that enable molecule visualization. Usually, it is a hard task to find the bestconditions to obtain high resolution images with low noise. An adequate STM image interpretationcan be affected by noise signals and therefore it is necessary to correct and eliminate them. Thetechnique needs a long-term temporal analysis (i.e. many hours) and the drift can be severe enoughto displace the scan window beyond the original regions of interest.This project intends to develop new algorithms to denoise the images acquired with STM. Theinnovation goal is to formulate the sparse denoise problem as a morphological component analysisproblem, where the components are related to the periodicity of use of the elements present in thelearned dictionary (periodic v.s. aperiodic). This can be solved recurring to recently proposed workon efficient computation of convolutional sparse representations.

Imaging inverse problems (IIPs) abound in the modern world. Medical imaging (CT, MRI, PET, ultrasound), remote sensing, seismography, non-destructive inspection, digital photography, astronomy, all involve at their computational core the solution of IIPs: they produce visual representations (images) of an underlying reality from indirect/imperfect observations. In a nutshell, the goal of this project is to advance the state of the art in computational methods for IIPs, by exploiting the promising new possibilities that lie outside of the classical convex optimization approaches. IIPs are ill-posed: even if the observation operator is perfectly known, the observations do not uniquely and stably determine the solution. This difficulty is dealt with by seeking a compromise between data fidelity (fitting the observed data) and adherence to properties that the unknown image is known/desired to have. The classical way to seek this balance is to formulate an optimization problem (usually convex, more by reasons of mathematical tractability than model adequacy), where the objective function includes a term encouraging data fidelity and another (the regularizer/prior) penalizing undesirable solutions; in this approach, convex optimization takes center stage and has been in the limelight of computational imaging.If the observation operator is only partially known (in the so-called blind problems), the goal is to infer, not only the underlying image, but also the missing information about the observation model. Blind problems are obviously more ill-posed, requiring regularizers/priors that better model the images of interest. Moreover, variational formulations of blind IIPs are naturally non-convex, thus out of reach for the efficient convex optimization tools developed in the past decade for the convex formulations of non-blind IIPs. Consequently, the methods that have been proposed for blind problems lack convergence guarantees and require "tricks" in order to work well (e.g...

The goal of the project is to advance the state of the art in criteria and algorithms for solving imaging inverse problems. The research fronts that we aim to push forward are the following: a) In standard photography, and even more in computational photography (CP), deconvolution plays a key role. In the standard case, deblurring is used to compensate lens softness, while in many CP applications (e.g., coded aperture) deconvolution is itself responsible for producing the images. Even with the CP ''friendly'' direct operators, prior knowledge (regularization) is required to solve the IP. In this work front, we will develop new regularization criteria, beyond the currently used sparseness-inducing 1-norm and total-variation. In particular, we will consider: compound regularizers; criteria that adapt automatically to the intrinsic complexity and/or the local structure of the underlying image; non-local regularization (which has only been used for pure denoising problems, with excellent results).b) Realistically, full knowledge about the convolution operator is seldom available: the exact motion causing a motion blur is of course unknown; coded aperture imaging is highly sensitive to the exact knowledge of the aperture, which is usually obtained by a previous careful calibration procedure. These observations stress the general importance of blind deconvolution. In this topic, we willresearch criteria for blind deconvolution, for problems with varying degrees of uncertainty about the direct operator. In particular, we will take steps towards the ''holly grail'' of blind deconvolution: an objective function whose minimization yields optimal image and convolution estimates.c) Some of the currently fastest algorithms for CS and image deconvolution under standard regularizers were developed by researchers in the project team. The new criteria mentioned in (a) and (b) will result in new, more challenging optimization problems, requiring new algorithms. This will be the third...

This project will address new, statistically based methods forimage restoration and reconstruction. The fundamental approach consists of usingrepresentations of natural images yielding sparse andnear-independent coefficients. This goal will be pursued along twomain lines:1. Methods relying on a fixed, wavelet-based representation. Theproject shall extend existing methods of denoising, and ofdeblurring using a known blur operator, to the blind deblurring /denoising situation. The main approach will consist of taking theblur operator as an additional unknown in a Bayesian formulationof the problem.2. Methods relying on a learned, ICA-based, spatially invariantrepresentation. The project shall address new image models,consisting of an i.i.d. source followed by a learned, linear ornonlinear, spatially invariant filter, and their application tothe blind deblurring problem.Besides blurred natural images, the project will also address thereconstruction of other classes of images, namely those formedthrough tomographic methods.