Search

The question of the correct formulation for the momentum of light in a dielectric medium is typically referred to as the ``Abraham-Minkowski controversy". Experiments conducted to elucidate the issue have primarily focused on measuring forces and momentum transfers. In this work, we propose an interferometric approach using matter waves to measure the light-induced version of the He-McKellar-Wilkens (optical HMW) phase for a neutral atomic dipole in dynamical electromagnetic fields. Beginning from the action principle, we show that this geometric phase is directly related via the Euler-Lagrange equations of motion to the Abraham force lying at the heart of the controversy., Comment: 6 pages, 2 figures

Recent years have seen a strongly increased visibility of non-binary people in public discourse. Accordingly, considerations of gender-fair language go beyond a binary conception of male/female. However, language technology, especially machine translation (MT), still suffers from binary gender bias. Proposing a solution for gender-fair MT beyond the binary from a purely technological perspective might fall short to accommodate different target user groups and in the worst case might lead to misgendering. To address this challenge, we propose a method and case study building on participatory action research to include experiential experts, i.e., queer and non-binary people, translators, and MT experts, in the MT design process. The case study focuses on German, where central findings are the importance of context dependency to avoid identity invalidation and a desire for customizable MT solutions., Comment: 11 pages, 4 figures

In principle, applying variational autoencoders (VAEs) to sequential data offers a method for controlled sequence generation, manipulation, and structured representation learning. However, training sequence VAEs is challenging: autoregressive decoders can often explain the data without utilizing the latent space, known as posterior collapse. To mitigate this, state-of-the-art models weaken the powerful decoder by applying uniformly random dropout to the decoder input. We show theoretically that this removes pointwise mutual information provided by the decoder input, which is compensated for by utilizing the latent space. We then propose an adversarial training strategy to achieve information-based stochastic dropout. Compared to uniform dropout on standard text benchmark datasets, our targeted approach increases both sequence modeling performance and the information captured in the latent space., Comment: Accepted at NeurIPS 2022

The accurate measurement of blood pressure (BP) is an important prerequisite for the reliable diagnosis and efficient management of hypertension and other medical conditions. Office Blood Pressure Measurement (OBP) is a technique performed in-office with the sphygmomanometer, while Ambulatory Blood Pressure Monitoring (ABPM) is a technique that measures blood pressure during 24h. The BP fluctuations also depend on other factors such as physical activity, temperature, mood, age, sex, any pathologies, a hormonal activity that may intrinsically influence the differences between OBP and ABPM. The aim of this study is to examine the possible influence of sex on the discrepancies between OBP and ABPM in 872 subjects with known or suspected hypertension. A significant correlation was observed between OBP and ABPM mean values calculated during the day, night and 24h (ABPMday, ABPMnight, ABPM24h) in both groups (p<0.0001). The main finding of this study is that no difference between sexes was observed in the relation between OBP and mean ABMP values except between systolic OBP and systolic ABPM during the night. In addition, this study showed a moderate correlation between BPs obtained with the two approaches with a great dispersion around the regression line which suggests that the two approaches cannot be used interchangeably.

Over the years motor deficit in Parkinson's Disease (PD) patients was largely studied, however, no consistent pattern of relations between quantitative electroencephalography (qEEG) and motor scales emerged. There is a general lack of information on the relation between EEG changes and scales related to specific motor deficits. Therefore, the study aimed to investigate the relation between brain oscillatory activity alterations (EEG power bands) and most used PD-related motor deficit scales. A positive correlation was found between the freezing of the gait questionnaire (FOGQ) and delta spectral power band (rho=0.67; p=0.008), while a negative correlation with the same scale was observed in the alpha spectral power band (rho=-0.59, p=0.027). Additionally, motor scores measure by motor part of Unified Parkinson's Disease Rating Scale (UPDRS) correlated directly with theta (rho=0.55, p=0.040) and inversely with beta EEG power band (rho=-0.77, p=0.001). No significant correlation was found between spectral powers and Hoehn and Yahr (H&Y), BERG (Berg K. et. al. 1995), Modified Parkinson Activity Scale (MPAS), Six-Minute Walk Test (6MWT) and Timed Up and Go Test (TUG). In conclusion, our study supports the earlier findings suggesting a link between EEG slowing and motor decline, providing more insight into the relation between EEG alteration and deficits in different motor domains. These findings indicate that EEG assessment may be a useful biomarker for objective monitoring of progression and neurophysiological effect of rehabilitation approaches in PD's.

In clinical practice, regions of interest in medical imaging often need to be identified through a process of precise image segmentation. The quality of this image segmentation step critically affects the subsequent clinical assessment of the patient status. To enable high accuracy, automatic image segmentation, we introduce a novel deep neural network architecture that exploits the inherent spatial coherence of anatomical structures and is well equipped to capture long-range spatial dependencies in the segmented pixel/voxel space. In contrast to the state-of-the-art solutions based on convolutional layers, our approach leverages on recently introduced spatial dependency layers that have an unbounded receptive field and explicitly model the inductive bias of spatial coherence. Our method performs favourably to commonly used U-Net and U-Net++ architectures as demonstrated by improved Dice and Jaccardscore in three different medical segmentation tasks: nuclei segmentation in microscopy images, polyp segmentation in colonoscopy videos, and liver segmentation in abdominal CT scans.

How to improve generative modeling by better exploiting spatial regularities and coherence in images? We introduce a novel neural network for building image generators (decoders) and apply it to variational autoencoders (VAEs). In our spatial dependency networks (SDNs), feature maps at each level of a deep neural net are computed in a spatially coherent way, using a sequential gating-based mechanism that distributes contextual information across 2-D space. We show that augmenting the decoder of a hierarchical VAE by spatial dependency layers considerably improves density estimation over baseline convolutional architectures and the state-of-the-art among the models within the same class. Furthermore, we demonstrate that SDN can be applied to large images by synthesizing samples of high quality and coherence. In a vanilla VAE setting, we find that a powerful SDN decoder also improves learning disentangled representations, indicating that neural architectures play an important role in this task. Our results suggest favoring spatial dependency over convolutional layers in various VAE settings. The accompanying source code is given at https://github.com/djordjemila/sdn.

AutoML systems provide a black-box solution to machine learning problems by selecting the right way of processing features, choosing an algorithm and tuning the hyperparameters of the entire pipeline. Although these systems perform well on many datasets, there is still a non-negligible number of datasets for which the one-shot solution produced by each particular system would provide sub-par performance. In this paper, we present Amazon SageMaker Autopilot: a fully managed system providing an automated ML solution that can be modified when needed. Given a tabular dataset and the target column name, Autopilot identifies the problem type, analyzes the data and produces a diverse set of complete ML pipelines including feature preprocessing and ML algorithms, which are tuned to generate a leaderboard of candidate models. In the scenario where the performance is not satisfactory, a data scientist is able to view and edit the proposed ML pipelines in order to infuse their expertise and business knowledge without having to revert to a fully manual solution. This paper describes the different components of Autopilot, emphasizing the infrastructure choices that allow scalability, high quality models, editable ML pipelines, consumption of artifacts of offline meta-learning, and a convenient integration with the entire SageMaker suite allowing these trained models to be used in a production setting.

Tuning complex machine learning systems is challenging. Machine learning typically requires to set hyperparameters, be it regularization, architecture, or optimization parameters, whose tuning is critical to achieve good predictive performance. To democratize access to machine learning systems, it is essential to automate the tuning. This paper presents Amazon SageMaker Automatic Model Tuning (AMT), a fully managed system for gradient-free optimization at scale. AMT finds the best version of a trained machine learning model by repeatedly evaluating it with different hyperparameter configurations. It leverages either random search or Bayesian optimization to choose the hyperparameter values resulting in the best model, as measured by the metric chosen by the user. AMT can be used with built-in algorithms, custom algorithms, and Amazon SageMaker pre-built containers for machine learning frameworks. We discuss the core functionality, system architecture, our design principles, and lessons learned. We also describe more advanced features of AMT, such as automated early stopping and warm-starting, showing in experiments their benefits to users.

Blood Pressure (BP) is a biological signal related to the cardiovascular system that inevitably is affected by ageing. Moreover, it is also influenced by the presence of cardiovascular risk factors. To evaluate how the relationship be-tween BP and age changes with the presence of risk factors in hypertensive and normotensive subjects, we analyzed 880 subjects with and without smoking, obe-sity, diabetes mellitus and dyslipidemia. A regression line fitted each BP/Age relation calculated separately for normotensive and hypertensive subjects with and without risk factors. For each of the four conditions the office and the 24-hour ambulatory BP monitoring (ABPM) were considered. In subjects with and without risk factors, the slopes of the Systolic BP/Age relation were higher in hypertensive than in normotensive subjects in both office and ABPM conditions. Moreover, the presence of risk factors modified the Systolic BP/Age relation in hypertensive subjects by using either office or ABPM measurements. Finally, we confirmed that the difference between the two modalities depends on age too.

Blood Pressure (BP) and Heart Rate (HR) provide information on clin-ical condition along 24h. Both signals present circadian changes due to sympa-thetic/parasympathetic control system that influence the relationship between them. Moreover, also the gender could modify this relation, acting on both con-trol systems. Some studies, using office measurements examined the BP/HR re-lation, highlighting a direct association between the two variables, linked to sus-pected coronary heart disease. Nevertheless, till now such relation has not been studied yet using ambulatory technique that is known to lead to additional prog-nostic information about cardiovascular risks. In order to examine in a more ac-curate way this relation, in this work we evaluate the influence of gender on the BP/HR relationship by using hour-to-hour 24h ambulatory measurements. Data coming from 122 female and 50 male normotensive subjects were recorded using a Holter Blood Pressure Monitor and the parameters of the linear regression fit-ting BP/HR were calculated. Results confirmed those obtained in previous stud-ies using punctual office measures in males and underlined a significant relation between Diastolic BP and HR during each hour of the day in females; a different trend in the BP/HR relation between genders was found only during night-time. Moreover, the circadian rhythm of BP/HR is similar in both genders but with different values of HR and BP at different times of the day.

The use of office measurement of Blood Pressure (BP) as well as of the mean on day-time, on night-time or on 24h does not accurately describe the changes of the BP circadian rhythm. Moreover, several risk factors affect this rhythm but until now possible alterations, due to the presence of such risk factors considered separately, were not been yet studied. Cigarette smoking is one of the most relevant risk factors increasing cardiovascular morbidity and mortality. The aim of this study is to evaluate quantitatively and with a suitable temporal detail how the smoking influences the BP circadian rhythm in normotensive and hypertensive subjects excluding those who presented other risk factors like obesity, dyslipidemia and diabetes mellitus. Holter BP monitoring coming from 618 subjects was used and the behaviour on 24h was examined separately in normotensive and hypertensive subjects either smokers or non-smokers. Four intervals with alternate different characteristics were found in the BP rhythm and regression lines approximated them in order to evaluate the changing rate of BP in each period. Results showed higher values from 10:00 to 02:00 in hypertensive smokers than non-smokers and significant differences between normotensive smokers and non-smokers between 10:00 and 19:00. The changing rate between 10:00 and 14:30 was higher in non-smokers than in smokers for both normotensive and hypertensive subjects while the opposite was found in the other three periods. The different velocity rates of BP changes during 24h, could be associated with different risk levels of cardiovascular disease., Comment: GNB conference

The study reports the performance of Parkinson's disease (PD) patients to operate Motor-Imagery based Brain-Computer Interface (MI-BCI) and compares three selected pre-processing and classification approaches. The experiment was conducted on 7 PD patients who performed a total of 14 MI-BCI sessions targeting lower extremities. EEG was recorded during the initial calibration phase of each session, and the specific BCI models were produced by using Spectrally weighted Common Spatial Patterns (SpecCSP), Source Power Comodulation (SPoC) and Filter-Bank Common Spatial Patterns (FBCSP) methods. The results showed that FBCSP outperformed SPoC in terms of accuracy, and both SPoC and SpecCSP in terms of the false-positive ratio. The study also demonstrates that PD patients were capable of operating MI-BCI, although with lower accuracy.

There is a growing research interest towards the use of wireless IMU sensors to assess disability, monitor progress and provide feedback to patients on range of motion and movement performance during upper body rehabilitation. The quality of movement in patients with adhesive capsulitis and relative treatment efficacy has not yet been studied using inertial and magnetic sensors. The aim of this study was to investigate the possibility to quantitatively evaluate capsulate-related deficit versus healthy controls and to assess treatment efficacy by measurement of shoulder kinematic parameters with ISEO protocol using inertial and magnetic measurement system technology. We enrolled 6 patients with adhesive capsulitis (AC) who underwent the experimental assessment by using a set of wireless IMU sensors at the baseline (T0) and after the 15 one-hour individual sessions of physiotherapy (T1). The range of motion in elevation, abduction and the scapulohumeral rhythm kinematic parameters were extracted from measurements performed in enrolled AC patients and in 7 healthy controls. The results of this preliminary study showed that proposed approach based on measurement of shoulder kinematic parameters with ISEO protocol using IMU wireless sensors can be useful in mobility deficit assessment of patients with adhesive capsulitis, as well as for monitoring of treatment efficacy and its further personalization.

Motor-Imagery based BCI (MI-BCI) neurorehabilitation can improve locomotor ability and reduce the deficit symptoms in Parkinson's Disease patients. Advanced Motor-Imagery BCI methods are needed to overcome the accuracy and time-related MI BCI calibration challenges in such patients. In this study, we proposed a Multi-session FBCSP (msFBCSP) based on inter-session transfer learning and we investigated its performance compared to the single-session based FBSCP. The main result of this study is the significantly improved accuracy obtained by proposed msFBCSP compared to single-session FBCSP in PD patients (median 81.3%, range 41.2-100.0% vs median 61.1%, range 25.0-100.0%, respectively; p<0.001). In conclusion, this study proposes a transfer learning-based multi-session based FBCSP approach which allowed to significantly improve calibration accuracy in MI BCI performed on PD patients.

The diagnosis of heart diseases is a difficult task generally addressed by an appropriate examination of patients clinical data. Recently, the use of heart rate variability (HRV) analysis as well as of some machine learning algorithms, has proved to be a valuable support in the diagnosis process. However, till now, ischemic heart disease (IHD) has been diagnosed on the basis of Artificial Neural Networks (ANN) applied only to signs, symptoms and sequential ECG and coronary angiography, an invasive tool, while could be probably identified in a non-invasive way by using parameters extracted from HRV, a signal easily obtained from the ECG. In this study, 18 non-invasive features (age, gender, left ventricular ejection fraction and 15 obtained from HRV) of 243 subjects (156 normal subjects and 87 IHD patients) were used to train and validate a series of several ANN, different for number of input and hidden nodes. The best result was obtained using 7 input parameters and 7 hidden nodes with an accuracy of 98.9% and 82% for the training and validation dataset, respectively.

Motor-imagery based brain-computer interfaces (MI-BCI) have the potential to become ground-breaking technologies for neurorehabilitation, the reestablishment of non-muscular communication and commands for patients suffering from neuronal disorders and disabilities, but also outside of clinical practice, for video game control and other entertainment purposes. However, due to the noisy nature of the used EEG signal, reliable BCI systems require specialized procedures for features optimization and extraction. This paper compares the two approaches, the Common Spatial Patterns with Linear Discriminant Analysis classifier (CSP-LDA), widely used in BCI for extracting features in Motor Imagery (MI) tasks, and the Dual-Augmented Lagrangian (DAL) framework with three different regularization methods: group sparsity with row groups (DAL-GLR), dual-spectrum (DAL-DS) and l1-norm regularization (DAL-L1). The test has been performed on 7 healthy subjects performing 5 BCI-MI sessions each. The preliminary results show that DAL-GLR method outperforms standard CSP-LDA, presenting 6.9% lower misclassification error (p-value = 0.008) and demonstrate the advantage of DAL framework for MI-BCI.

This article starts from the assumption that near future 100BTransistor SuperComputers-on-a-Chip will include N big multi-core processors, 1000N small many-core processors, a TPU-like fixed-structure systolic array accelerator for the most frequently used Machine Learning algorithms needed in bandwidth-bound applications and a flexible-structure reprogrammable accelerator for less frequently used Machine Learning algorithms needed in latency-critical applications.