Search

Some applications of deep learning require not only to provide accurate results but also to quantify the amount of confidence in their prediction. The management of an electric power grid is one of these cases: to avoid risky scenarios, decision-makers need both precise and reliable forecasts of, for example, power loads. For this reason, point forecasts are not enough hence it is necessary to adopt methods that provide an uncertainty quantification. This work focuses on reservoir computing as the core time series forecasting method, due to its computational efficiency and effectiveness in predicting time series. While the reservoir computing literature mostly focused on point forecasting, this work explores the compatibility of some popular uncertainty quantification methods with the reservoir setting. Both Bayesian and deterministic approaches to uncertainty assessment are evaluated and compared in terms of their prediction accuracy, computational resource efficiency and reliability of the estimated uncertainty, based on a set of carefully chosen performance metrics.

In the field of earth observation (EO), continual learning (CL) algorithms have been proposed to deal with large datasets by decomposing them into several subsets and processing them incrementally. The majority of these algorithms assume that data are, first, coming from a single source, and second, fully labeled. Real-world EO datasets are instead characterized by a large heterogeneity (e.g., coming from aerial, satellite, or drone scenarios), and for the most part they are unlabeled, meaning they can be fully exploited only through the emerging self-supervised learning (SSL) paradigm. For these reasons, in this article, we present a new algorithm for merging SSL and CL for remote sensing applications that we call continual Barlow twins. It combines the advantages of one of the simplest self-supervision techniques, i.e., Barlow twins, with the elastic weight consolidation method to avoid catastrophic forgetting. In addition, we evaluate the proposed continual SSL approach on a highly heterogeneous EO dataset, showing the effectiveness of this strategy on a novel combination of three almost non-overlapping domains datasets (airborne Potsdam, satellite US3D, and drone unmanned aerial vehicle semantic segmentation dataset), on a crucial downstream task in EO, i.e., semantic segmentation. Encouraging results show the superiority of SSL in this setting, and the effectiveness of creating an incremental effective pretrained feature extractor, based on ResNet50, without the need of relying on the complete availability of all the data, with a valuable saving of time and resources.

Recent research has found that neural networks for computer vision are vulnerable to several types of external attacks that modify the input of the model, with the malicious intent of producing a misclassification. With the increase in the number of feasible attacks, many defence approaches have been proposed to mitigate the effect of these attacks and protect the models. Mainly, the research on both attack and defence has focused on RGB images, while other domains, such as the infrared domain, are currently underexplored. In this paper, we propose two attacks, and we evaluate them on multiple datasets and neural network models, showing that the results outperform others established attacks, on both RGB as well as infrared domains. In addition, we show that our proposal can be used in an adversarial training protocol to produce more robust models, with respect to both adversarial attacks and natural perturbations that can be applied to input images. Lastly, we study if a successful attack in a domain can be transferred to an aligned image in another domain, without any further tuning. The code, containing all the files and the configurations used to run the experiments, is available https://github.com/jaryP/IR-RGB-domain-attackonline.

Over the last 20 years we have seen an increase in techniques in the field of computational pathology and machine learning, improving our ability to analyze and interpret imaging. Neural networks, in particular, have been used for more than thirty years, starting with the computer assisted smear test using early generation models. Today, advanced machine learning, working on large image data sets, has been shown to perform classification, detection, and segmentation with remarkable accuracy and generalization in several domains. Deep learning algorithms, as a branch of machine learning, are thus attracting attention in digital pathology and cytopathology, providing feasible solutions for accurate and efficient cytological diagnoses, ranging from efficient cell counts to automatic classification of anomalous cells and queries over large clinical databases. The integration of machine learning with related next-generation technologies powered by AI, such as augmented/virtual reality, metaverse, and computational linguistic models are a focus of interest in health care digitalization, to support education, diagnosis, and therapy. In this work we will consider how all these innovations can help cytopathology to go beyond the microscope and to undergo a hyper-digitalized transformation. We also discuss specific challenges to their applications in the field, notably, the requirement for large-scale cytopathology datasets, the necessity of new protocols for sharing information, and the need for further technological training for pathologists.

BackgroundHandwriting is an acquired complex cognitive and motor skill resulting from the activation of a widespread brain network. Handwriting therefore may provide biologically relevant information on health status. Also, handwriting can be collected easily in an ecological scenario, through safe, cheap, and largely available tools. Hence, objective handwriting analysis through artificial intelligence would represent an innovative strategy for telemedicine purposes in healthy subjects and people affected by neurological disorders.Materials and MethodsOne-hundred and fifty-six healthy subjects (61 males; 49.6 ± 20.4 years) were enrolled and divided according to age into three subgroups: Younger adults (YA), middle-aged adults (MA), and older adults (OA). Participants performed an ecological handwriting task that was digitalized through smartphones. Data underwent the DBNet algorithm for measuring and comparing the average stroke sizes in the three groups. A convolutional neural network (CNN) was also used to classify handwriting samples. Lastly, receiver operating characteristic (ROC) curves and sensitivity, specificity, positive, negative predictive values (PPV, NPV), accuracy and area under the curve (AUC) were calculated to report the performance of the algorithm.ResultsStroke sizes were significantly smaller in OA than in MA and YA. The CNN classifier objectively discriminated YA vs. OA (sensitivity = 82%, specificity = 80%, PPV = 78%, NPV = 79%, accuracy = 77%, and AUC = 0.84), MA vs. OA (sensitivity = 84%, specificity = 56%, PPV = 78%, NPV = 73%, accuracy = 74%, and AUC = 0.7), and YA vs. MA (sensitivity = 75%, specificity = 82%, PPV = 79%, NPV = 83%, accuracy = 79%, and AUC = 0.83).DiscussionHandwriting progressively declines with human aging. The effect of physiological aging on handwriting abilities can be detected remotely and objectively by using machine learning algorithms.

In recent years, hyper-complex deep networks (such as complex-valued and quaternion-valued neural networks – QVNNs) have received a renewed interest in the literature. They find applications in multiple fields, ranging from image reconstruction to 3D audio processing. Similar to their real-valued counterparts, quaternion neural networks require custom regularisation strategies to avoid overfitting. In addition, for many real-world applications and embedded implementations, there is the need of designing sufficiently compact networks, with few weights and neurons. However, the problem of regularising and/or sparsifying QVNNs has not been properly addressed in the literature as of now. In this study, the authors show how to address both problems by designing targeted regularisation strategies, which can minimise the number of connections and neurons of the network during training. To this end, they investigate two extensions of [inline-formula] and structured regularisations to the quaternion domain. In the authors’ experimental evaluation, they show that these tailored strategies significantly outperform classical (real-valued) regularisation approaches, resulting in small networks especially suitable for low-power and real-time applications.

In this paper, we propose a new approach to train a deep neural network with multiple intermediate auxiliary classifiers, branching from it. These ‘multi-exits’ models can be used to reduce the inference time by performing early exit on the intermediate branches, if the confidence of the prediction is higher than a threshold. They rely on the assumption that not all the samples require the same amount of processing to yield a good prediction. In this paper, we propose a way to train jointly all the branches of a multi-exit model without hyper-parameters, by weighting the predictions from each branch with a trained confidence score. Each confidence score is an approximation of the real one produced by the branch, and it is calculated and regularized while training the rest of the model. We evaluate our proposal on a set of image classification benchmarks, using different neural models and early-exit stopping criteria.

Scene understanding of satellite and aerial images is a pivotal task in various remote sensing (RS) practices, such as land cover and urban development monitoring. In recent years, neural networks have become a de-facto standard in many of these applications. However, semantic segmentation still remains a challenging task. With respect to other computer vision (CV) areas, in RS large labeled datasets are not very often available, due to their large cost and to the required manpower. On the other hand, self-supervised learning (SSL) is earning more and more interest in CV, reaching state-of-the-art in several tasks. In spite of this, most SSL models, pretrained on huge datasets like ImageNet, do not perform particularly well on RS data. For this reason, we propose a combination of a SSL algorithm (particularly, Online Bag of Words) and a semantic segmentation algorithm, shaped for aerial images (namely, Multistage Attention ResU-Net), to show new encouraging results (i.e., 81.76% mIoU with ResNet-18 backbone) on the ISPRS Vaihingen dataset.