Abstract
In order to increase the adoption of machine learning models in areas like medicine and finance, it is necessary to have correct and diverse explanations for the decisions that the models provide, to satisfy the curiosity of decision-makers and the needs of the regulators. In this paper, we introduced a method, based in a previously presented framework, to explain the decisions of an Ensemble Model. Moreover, we instantiate the proposed approach to an ensemble composed of a Scorecard, a Random Forest, and a Deep Neural Network, to produce accurate decisions along with correct and diverse explanations. Our methods are tested on two biomedical datasets and one financial dataset. The proposed ensemble leads to an improvement in the quality of the decisions, and in the correctness of the explanations, when compared to its constituents alone. Qualitatively, it produces diverse explanations that make sense and convince the experts.

Abstract
Overhead Power Distribution Lines (OPDLs) correspond to a large percentage of the medium-voltage electrical systems. In these networks, visual inspection activities are usually performed without resorting to automated systems, requiring a significant investment of time and human resources. We present a methodology to identify the defect and type of insulators using Convolutional Neural Networks (CNNs). More than 2500 photographs were collected both from inside a studio and from a realistic OPDL. A classification model is proposed to automatically recognize the insulators conformity. This model is able to learn from indoors photographs by augmenting these images with realistic details such as top ties and real-world backgrounds. Furthermore, Multi-Task Learning (MTL) was used to improve performance of defect detection by also predicting the insulator class. The proposed methodology is able to achieve an accuracy of 92% for material classification and 85% for defect detection, with F1-score of 0.75, surpassing available solutions.

Abstract
Semantic segmentation consists in predicting whether any given pixel is part of the object of interest or not. Two types of errors are therefore possible: false positives and false negatives. For visualization and emphasis purposes, we might want to put special effort into reducing one type of error in detriment of the other. A common practice is to define the two types of errors as a relative trade-off using a cost matrix. However, it might be more natural for humans to define the trade-off in terms of an absolute constraint on one type of errors while trying to minimize the other. Previously, we suggested possible approaches to introduce this absolute trade-off in binary classifiers. Extending to semantic segmentation, we propose a threshold on the sigmoid layer and modifications to gradient descent such as adding a new term to the loss function and training in two phases. The latter produced the more resilient results, with a simple threshold being sufficient in most cases.

Abstract
The segmentation of blood vessels in medical images has been heavily studied, given its impact in several clinical practices. Deep Learning methods have been applied to supervised segmentation of blood vessels, mainly the retinal ones due to the availability of manual annotations. Despite their success, they typically minimize the Binary Cross Entropy loss, which does not penalize topological mistakes. These errors are relevant in graph-like structures such as blood vessel trees, as a missing segment or an inadequate merging or splitting of branches, may severely change the topology of the network and put at risk the extraction of vessel pathways and their characterization. In this paper, we propose an end-to-end network design comprising a cascade of a typical segmentation network and a Variational Auto-Encoder which, by learning a rich but compact latent space, is able to correct many topological incoherences. Our experiments in three of the most commonly used retinal databases, DRIVE, STARE, and CHASEDB1, show that the proposed model effectively learns representations inducing better segmentations in terms of topology, without hurting the usual pixel-wise metrics.

Abstract
Convolutional Neural Networks (CNN) have become the gold standard in many visual recognition tasks including medical applications. Due to their high variance, however, these models are prone to over-fit the data they are trained on. To mitigate this problem, one of the most common strategies, is to perform data augmentation. Rotation, scaling and translation are common operations. In this work we propose an alternative method to rotation-based data augmentation where the rotation transformation is performed inside the CNN architecture. In each training batch the weights of all convolutional layers are rotated by the same random angle. We validate our proposed method empirically showing its usefulness under different scenarios.

Abstract
We propose a framework to model the distribution of sequential data coming from a set of entities connected in a graph with a known topology. The method is based on a mixture of shared hidden Markov models (HMMs), which are jointly trained in order to exploit the knowledge of the graph structure and in such a way that the obtained mixtures tend to be sparse. Experiments in different application domains demonstrate the effectiveness and versatility of the method.