Search

Information about Cu fractionation and Cu isotopic composition can be paramount when investigating Wilson's disease (WD). This information can provide a better understanding of the metabolism of Cu. Most importantly, it may provide an easy way to diagnose and to follow the evolution of WD patients. For such purposes, protocols for Cu determination and Cu isotopic analysis via inductively coupled plasma mass spectrometry were investigated in this work, both in bulk serum and in the exchangeable copper (CuEXC) fractions. The CuEXC protocol provided satisfactory recovery values. Also, no significant mass fractionation during the whole analytical procedure (CuEXC production and/or Cu isolation) was detected. Analyses were carried out in controls (healthy persons), newborns, patients with hepatic disorders, and WD patients. While the results for Cu isotopic analysis are relevant (e.g., δ65Cu values were lower for both WD patients under chelating treatment and patients with hepatic problems in comparison with those values obtained for WD patients under Zn treatments, controls, and newborns) to comprehend Cu metabolism and to follow up the disease, the parameter that can help to better discern between WD patients and the rest of the patients tested (non-WD) was found to be the REC (relative exchangeable Cu). In this study, all the WD patients showed a REC higher than 17%, while the rest showed lower values. However, since establishing a universal threshold is complicated, machine learning was investigated to produce a model that can differentiate between WD and non-WD samples with excellent results (100% accuracy, albeit for a limited sample set). Most importantly, unlike other ML approaches, our model can also provide an uncertainty metric to indicate the reliability of the prediction, overall opening new ways to diagnose WD.

Convolutional neural networks (CNNs) have proven to be very efficient for the analysis of remote sensing (RS) images. Due to the inherent complexity of extracting features from these images, along with the increasing amount of data to be processed (and the diversity of applications), there is a clear tendency to develop and employ increasingly deep and complex CNNs. In this regard, graphics processing units (GPUs) are frequently used to optimize their execution, both for the training and inference stages, optimizing the performance of neural models through their many-core architecture. Hence, the efficient use of the GPU resources should be at the core of optimizations. This letter analyzes the possibilities of using a new family of CNNs, denoted as TResNets, to provide an efficient solution to the RS scene classification problem. Moreover, the considered models have been combined with mixed precision to enhance their training performance. Our experimental results, conducted over three publicly available RS data sets, show that the proposed networks achieve better accuracy and more efficient use of GPU resources than other state-of-the-art networks. Source code is available at https://github.com/mhaut/GPUfriendlyRS.

Machine learning techniques, and specifically neural networks, have proved to be very useful tools for image classification tasks. Nevertheless, measuring the reliability of these networks and calibrating them accurately are very complex. This is even more complex in a field like hyperspectral imaging, where labeled data are scarce and difficult to generate. Bayesian neural networks (BNNs) allow to obtain uncertainty metrics related to the data processed (aleatoric), and to the uncertainty generated by the model selected (epistemic). On this work, we will demonstrate the utility of BNNs by analyzing the uncertainty metrics obtained by a BNN over five of the most used hyperspectral images datasets. In addition, we will illustrate how these metrics can be used for several practical applications such as identifying predictions that do not reach the required level of accuracy, detecting mislabeling in the dataset, or identifying when the predictions are affected by the increase of the level of noise in the input data.

This paper presents a novel approach, based on the standard addition method, for overcoming the matrix effects that often hamper the accurate characterization of nanoparticles (NPs) in complex samples via single particle inductively coupled plasma mass spectrometry (SP-ICP-MS). In this approach, calibration of the particle size is performed by two different methods: (i) by spiking a suspension of NPs standards of known size containing the analyte, or (ii) by spiking the sample with ionic standards; either way, the measured sensitivity is used in combination with the transport efficiency (TE) for sizing the NPs. Moreover, such transport efficiency can be readily obtained from the data obtained via both calibration methods mentioned above, so that the particle number concentration can also be determined. The addition of both ionic and NP standards can be performed on-line, by using a T-piece with two inlet lines of different dimensions. The smaller of the two is used for the standards, thus ensuring a constant and minimal sample dilution. As a result of the spiking of the samples, mixed histograms including the signal of the sample and that of the standards are obtained. However, the use of signal deconvolution approaches permits to extract the information, even in cases of signal populations overlapping. For proofing the concept, characterization of a 50 nm AuNPs suspension prepared in three different media (i.e., deionized water, 5% ethanol, and 2.5% tetramethyl ammonium hydroxide-TMAH) was carried out. Accurate results were obtained in all cases, in spite of the matrix effects detected in some media. Overall, the approach proposed offers flexibility, so it can be adapted to different situations, but it might be specially indicated for samples for which the matrix is not fully known and/or dilution is not possible/recommended.

A decision tree is a well-known machine learning technique. Recently their popularity has increased due to the powerful Gradient Boosting ensemble method that allows to gradually increasing accuracy at the cost of executing a large number of decision trees. In this paper we present an accelerator designed to optimize the execution of these trees while reducing the energy consumption. We have implemented it in an FPGA for embedded systems, and we have tested it with a relevant case-study: pixel classification of hyperspectral images. In our experiments with different images our accelerator can process the hyperspectral images at the same speed at which they are generated by the hyperspectral sensors. Compared to a high-performance processor running optimized software, on average our design is twice as fast and consumes 72 times less energy. Compared to an embedded processor, it is 30 times faster and consumes 23 times less energy.

Estimatabion of wave agitation plays a key role in predicting natural disasters, path optimization and secure harbor operation. The Spanish agency Puertos del Estado (PdE) has several oceanographic measure networks equipped with sensors for different physical variables, and manages forecast systems involving numerical models. In recent years, there is a growing interest in wave parameter estimation by using machine learning models due to the large amount of oceanographic data available for training, as well as its proven efficacy in estimating physical variables. In this study, we propose to use machine learning techniques to improve the accuracy of the current forecast system of PdE. We have focused on four physical wave variables: spectral significant height, mean spectral period, peak period and mean direction of origin. Two different machine learning models have been explored: multilayer perceptron and gradient boosting decision trees, as well as ensemble methods that combine both models. These models reduce the error of the predictions of the numerical model by 36% on average, demonstrating the potential gains of combining machine learning and numerical models.

Deep neural networks (DNNs) are increasing their presence in a wide range of applications, and their computationally intensive and memory-demanding nature poses challenges, especially for embedded systems. Pruning techniques turn DNN models into sparse by setting most weights to zero, offering optimization opportunities if specific support is included. We propose a novel pipelined architecture for DNNs that avoids all useless operations during the inference process. It has been implemented in a field-programmable gate array (FPGA), and the performance, energy efficiency, and area have been characterized. Exploiting sparsity yields remarkable speedups but also produces area overheads. We have evaluated this tradeoff in order to identify in which scenarios it is better to use that area to exploit sparsity, or to include more computational resources in a conventional DNN architecture. We have also explored different arithmetic bitwidths. Our sparse architecture is clearly superior on 32-bit arithmetic or highly sparse networks. However, on 8-bit arithmetic or networks with low sparsity it is more profitable to deploy a dense architecture with more arithmetic resources than including support for sparsity. We consider that FPGAs are the natural target for DNN sparse accelerators since they can be loaded at run-time with the best-fitting accelerator.

Board games applications usually offer a great user experience when running on desktop computers. Powerful high-performance processors working without energy restrictions successfully deal with the exploration of large game trees, delivering strong play to satisfy demanding users. However, nowadays, more and more game players are running these games on smartphones and tablets, where the lower computational power and limited power budget yield a much weaker play. Recent systems-on-a-chip include programmable logic tightly coupled with general-purpose processors enabling the inclusion of custom accelerators for any application to improve both performance and energy efficiency. In this paper, we analyze the benefits of partitioning the artificial intelligence of board games into software and hardware. We have chosen as case studies three popular and complex board games, Reversi , Blokus , and Connect6 . The designs analyzed include hardware accelerators for board processing, which improve performance and energy efficiency by an order of magnitude leading to much stronger and battery-aware applications. The results demonstrate that the use of hardware/software codesign to develop board games allows sustaining or even improving the user experience across platforms while keeping power and energy low.