Abstract
Decision trees are often preferred when implementing Machine Learning in embedded systems for their simplicity and scalability. Hoeffding Trees are a type of Decision Trees that take advantage of the Hoeffding Bound to allow them to learn patterns in data without having to continuously store the data samples for future reprocessing. This makes them especially suitable for deployment on embedded devices. In this work we highlight the features of a HLS implementation of the Hoeffding Tree. The implementation parameters include the feature size of the samples (D), the number of output classes (K), and the maximum number of nodes to which the tree is allowed to grow (Nd). We target a Xilinx MPSoC ZCU102, and evaluate: the design's resource requirements and clock frequency for different numbers of classes and feature size, the execution time on several synthetic datasets of varying sizes (N) and the execution time and accuracy for two datasets from UCI. For a problem size of D=3, K=5, and N=40000, a single decision tree operating at 103MHz is capable of 8.3x faster inference than the 1.2 GHz ARM Cortex-A53 core. Compared to a reference implementation of the Hoeffding tree, we achieve comparable classification accuracy for the UCI datasets.

Abstract
Energy efficiency is a non-functional requirement that developers must consider. This requirement is particularly relevant when building software for battery-operated devices like mobile ones: a long-lasting battery is an essential requirement for an enjoyable user experience. It has been shown that many mobile applications include inefficiencies that cause battery to be drained faster than necessary. Some of these inefficiencies result from software patterns that have been catalogued in the literature. The catalogues often provide more energy-efficient alternatives. While the related literature is vast, most approaches so far assume as a fundamental requirement that one has access to the source code of an application in order to be able to analyse it. This requirement makes independent energy analysis challenging, or even impossible, e.g. for a mobile user or, most significantly, an App Store trying to provide information on how efficient an application being submitted for publication is. Our work studies the viability of looking for known energy patterns in applications by decompiling them and analysing the resulting code. For this, we decompiled and analysed 236 open-source applications. We extended an existing tool to aid in this process, making it capable of seamlessly decompiling and analysing android applications. With the collected data, we performed a comparative analysis of the presence of energy patterns between the source code and the decompiled code. While further research is required to more assertively say if this type of static analysis is viable, our results point in a promising direction with 163 applications, approximately 69%, containing the same number of detected patterns in both source code and the release APK.

Abstract

Abstract
This study aimed to validate the Patient Health Questionnaire (PHQ-8) in a sample of military personnel, through the analysis of psychometric properties, reliability, and confirmatory factorial analysis. The questionnaire consists of 8 items that allow the assessment of depressive symptoms. The sample included 127 Portuguese military personnel aged between 21 and 78 years old. The results revealed a good internal consistency (alpha=.90) and good adjustment indices (chi 2/df=1.332, GFI=.956, CFI=.988, RMSEA=.051, SRMR=.30). In addition, convergent validity also showed to be good and composite reliability was .873. Thus, the PHQ-8 reveals good psychometric properties, being recommended for use in clinical practice and research with Portuguese military.

Abstract

Abstract