Abstract: Federated learning enables resource-constrained edge compute devices, such as mobile phones and IoT devices, to learn a shared model for prediction, while keeping the training data local. This decentralized approach to train models provides privacy, security, regulatory and economic benefits. In this work, we focus on the statistical challenge of federated learning when local data is non-IID. We first show that the accuracy of federated learning reduces significantly, by up to 55% for neural networks trained for highly skewed non-IID data, where each client device trains only on a single class of data. We further show that this accuracy reduction can be explained by the weight divergence, which can be quantified by the earth mover's distance (EMD) between the distribution over classes on each device and the population distribution. As a solution, we propose a strategy to improve training on non-IID data by creating a small subset of data which is globally shared between all the edge devices. Experiments show that accuracy can be increased by 30% for the CIFAR-10 dataset with only 5% globally shared data.

Abstract: On December 12th 2019, a new coronavirus (SARS-Cov2) emerged in Wuhan, China, sparking a pandemic of acute respiratory syndrome in humans (COVID-19). On the 24th of April 2020, the number of COVID-19 deaths in the world, according to the COVID-Case Tracker by Johns Hopkins University, was 195,313, and the number of COVID-19 confirmed cases was 2,783,512. The COVID-19 pandemic represents a massive impact on human health, causing sudden lifestyle changes, through social distancing and isolation at home, with social and economic consequences. Optimizing public health during this pandemic requires not only knowledge from the medical and biological sciences, but also of all human sciences related to lifestyle, social and behavioural studies, including dietary habits and lifestyle. Our study aimed to investigate the immediate impact of the COVID-19 pandemic on eating habits and lifestyle changes among the Italian population aged ≥ 12 years. The study comprised a structured questionnaire packet that inquired demographic information (age, gender, place of residence, current employment); anthropometric data (reported weight and height); dietary habits information (adherence to the Mediterranean diet, daily intake of certain foods, food frequency, and number of meals/day); lifestyle habits information (grocery shopping, habit of smoking, sleep quality and physical activity). The survey was conducted from the 5th to the 24th of April 2020. A total of 3533 respondents have been included in the study, aged between 12 and 86 years (76.1% females). The perception of weight gain was observed in 48.6% of the population; 3.3% of smokers decided to quit smoking; a slight increased physical activity has been reported, especially for bodyweight training, in 38.3% of respondents; the population group aged 18–30 years resulted in having a higher adherence to the Mediterranean diet when compared to the younger and the elderly population (p < 0.001; p < 0.001, respectively); 15% of respondents turned to farmers or organic, purchasing fruits and vegetables, especially in the North and Center of Italy, where BMI values were lower. In this study, we have provided for the first time data on the Italian population lifestyle, eating habits and adherence to the Mediterranean Diet pattern during the COVID-19 lockdown. However, as the COVID-19 pandemic is ongoing, our data need to be confirmed and investigated in future more extensive population studies.

Abstract: Autism spectrum disorder (ASD) is a highly heritable and heterogeneous group of neurodevelopmental phenotypes diagnosed in more than 1% of children. Common genetic variants contribute substantially to ASD susceptibility, but to date no individual variants have been robustly associated with ASD. With a marked sample-size increase from a unique Danish population resource, we report a genome-wide association meta-analysis of 18,381 individuals with ASD and 27,969 controls that identified five genome-wide-significant loci. Leveraging GWAS results from three phenotypes with significantly overlapping genetic architectures (schizophrenia, major depression, and educational attainment), we identified seven additional loci shared with other traits at equally strict significance levels. Dissecting the polygenic architecture, we found both quantitative and qualitative polygenic heterogeneity across ASD subtypes. These results highlight biological insights, particularly relating to neuronal function and corticogenesis, and establish that GWAS performed at scale will be much more productive in the near term in ASD.

Abstract: BackgroundSpinal muscular atrophy is an autosomal recessive neuromuscular disorder that is caused by an insufficient level of survival motor neuron (SMN) protein. Nusinersen is an antisense oligonu...

Abstract: Developing neural network image classification models often requires significant architecture engineering. In this paper, we study a method to learn the model architectures directly on the dataset of interest. As this approach is expensive when the dataset is large, we propose to search for an architectural building block on a small dataset and then transfer the block to a larger dataset. The key contribution of this work is the design of a new search space (the "NASNet search space") which enables transferability. In our experiments, we search for the best convolutional layer (or "cell") on the CIFAR-10 dataset and then apply this cell to the ImageNet dataset by stacking together more copies of this cell, each with their own parameters to design a convolutional architecture, named "NASNet architecture". We also introduce a new regularization technique called ScheduledDropPath that significantly improves generalization in the NASNet models. On CIFAR-10 itself, NASNet achieves 2.4% error rate, which is state-of-the-art. On ImageNet, NASNet achieves, among the published works, state-of-the-art accuracy of 82.7% top-1 and 96.2% top-5 on ImageNet. Our model is 1.2% better in top-1 accuracy than the best human-invented architectures while having 9 billion fewer FLOPS - a reduction of 28% in computational demand from the previous state-of-the-art model. When evaluated at different levels of computational cost, accuracies of NASNets exceed those of the state-of-the-art human-designed models. For instance, a small version of NASNet also achieves 74% top-1 accuracy, which is 3.1% better than equivalently-sized, state-of-the-art models for mobile platforms. Finally, the learned features by NASNet used with the Faster-RCNN framework surpass state-of-the-art by 4.0% achieving 43.1% mAP on the COCO dataset.

Abstract: Multiple myeloma (MM) is characterized by an uninhibited, clonal growth of plasma cells. While first-degree relatives of patients with MM show an increased risk of MM, the genetic basis of inherited MM susceptibility is incompletely understood. Here we report a genome-wide association study in the Nordic region identifying a novel MM risk locus at ELL2 (rs56219066T; odds ratio (OR)=1.25; P=9.6 × 10(-10)). This gene encodes a stoichiometrically limiting component of the super-elongation complex that drives secretory-specific immunoglobulin mRNA production and transcriptional regulation in plasma cells. We find that the MM risk allele harbours a Thr298Ala missense variant in an ELL2 domain required for transcription elongation. Consistent with a hypomorphic effect, we find that the MM risk allele also associates with reduced levels of immunoglobulin A (IgA) and G (IgG) in healthy subjects (P=8.6 × 10(-9) and P=6.4 × 10(-3), respectively) and, potentially, with an increased risk of bacterial meningitis (OR=1.30; P=0.0024).

Abstract: Starting from the spring of 2020, the outbreak of the COVID-19 caused Chinese universities to close the campuses and forced them to initiate online teaching. This paper focuses on a case of Peking University's online education. Six specific instructional strategies are presented to summarize current online teaching experiences for university instructors who might conduct online education in similar circumstances. The study concludes with five high-impact principles for online education: (a) high relevance between online instructional design and student learning, (b) effective delivery on online instructional information, (c) adequate support provided by faculty and teaching assistants to students; (d) high-quality participation to improve the breadth and depth of student's learning, and (e) contingency plan to deal with unexpected incidents of online education platforms.

Abstract: Unlike human learning, machine learning often fails to handle changes between training (source) and test (target) input distributions. Such domain shifts, common in practical scenarios, severely damage the performance of conventional machine learning methods. Supervised domain adaptation methods have been proposed for the case when the target data have labels, including some that perform very well despite being "frustratingly easy" to implement. However, in practice, the target domain is often unlabeled, requiring unsupervised adaptation. We propose a simple, effective, and efficient method for unsupervised domain adaptation called CORrelation ALignment (CORAL). CORAL minimizes domain shift by aligning the second-order statistics of source and target distributions, without requiring any target labels. Even though it is extraordinarily simple–it can be implemented in four lines of Matlab code–CORAL performs remarkably well in extensive evaluations on standard benchmark datasets.

Abstract: Background Transthyretin amyloid cardiomyopathy is caused by the deposition of transthyretin amyloid fibrils in the myocardium. The deposition occurs when wild-type or variant transthyretin becomes unstable and misfolds. Tafamidis binds to transthyretin, preventing tetramer dissociation and amyloidogenesis. Methods In a multicenter, international, double-blind, placebo-controlled, phase 3 trial, we randomly assigned 441 patients with transthyretin amyloid cardiomyopathy in a 2:1:2 ratio to receive 80 mg of tafamidis, 20 mg of tafamidis, or placebo for 30 months. In the primary analysis, we hierarchically assessed all-cause mortality, followed by frequency of cardiovascular-related hospitalizations according to the Finkelstein–Schoenfeld method. Key secondary end points were the change from baseline to month 30 for the 6-minute walk test and the score on the Kansas City Cardiomyopathy Questionnaire–Overall Summary (KCCQ-OS), in which higher scores indicate better health status. Results In the prim...

Abstract: We derive a new exact evolution equation for the scale dependence of an effective action. The corresponding equation for the effective potential permits a useful truncation. This allows one to deal with the infrared problems of theories with massless modes in less than four dimensions which are relevant for the high temperature phase transition in particle physics or the computation of critical exponents in statistical mechanics.

Abstract: Quantum spin liquids may be considered "quantum disordered" ground states of spin systems, in which zero point fluctuations are so strong that they prevent conventional magnetic long range order. More interestingly, quantum spin liquids are prototypical examples of ground states with massive many-body entanglement, of a degree sufficient to render these states distinct phases of matter. Their highly entangled nature imbues quantum spin liquids with unique physical aspects, such as non-local excitations, topological properties, and more. In this review, we discuss the nature of such phases and their properties based on paradigmatic models and general arguments, and introduce theoretical technology such as gauge theory and partons that are conveniently used in the study of quantum spin liquids. An overview is given of the different types of quantum spin liquids and the models and theories used to describe them. We also provide a guide to the current status of experiments to study quantum spin liquids, and to the diverse probes used therein.

Abstract: PurposeNivolumab provides clinical benefit (objective response rate [ORR], 31%; 95% CI, 20.8 to 42.9; disease control rate, 69%; 12-month overall survival [OS], 73%) in previously treated patients with DNA mismatch repair–deficient (dMMR)/microsatellite instability–high (MSI-H) metastatic colorectal cancer (mCRC); nivolumab plus ipilimumab may improve these outcomes. Efficacy and safety results for the nivolumab plus ipilimumab cohort of CheckMate-142, the largest single-study report of an immunotherapy combination in dMMR/MSI-H mCRC, are reported.Patients and MethodsPatients received nivolumab 3 mg/kg plus ipilimumab 1 mg/kg once every 3 weeks (four doses) followed by nivolumab 3 mg/kg once every 2 weeks. Primary end point was investigator-assessed ORR.ResultsOf 119 patients, 76% had received ≥ two prior systemic therapies. At median follow-up of 13.4 months, investigator-assessed ORR was 55% (95% CI, 45.2 to 63.8), and disease control rate for ≥ 12 weeks was 80%. Median duration of response was not reac...

Abstract: Atomically dispersed noble metal catalysts often exhibit high catalytic performances, but the metal loading density must be kept low (usually below 0.5%) to avoid the formation of metal nanoparticles through sintering. We report a photochemical strategy to fabricate a stable atomically dispersed palladium-titanium oxide catalyst (Pd1/TiO2) on ethylene glycolate (EG)-stabilized ultrathin TiO2 nanosheets containing Pd up to 1.5%. The Pd1/TiO2 catalyst exhibited high catalytic activity in hydrogenation of C=C bonds, exceeding that of surface Pd atoms on commercial Pd catalysts by a factor of 9. No decay in the activity was observed for 20 cycles. More important, the Pd1/TiO2-EG system could activate H2 in a heterolytic pathway, leading to a catalytic enhancement in hydrogenation of aldehydes by a factor of more than 55.

Abstract: It is known that the learning rate is the most important hyper-parameter to tune for training deep neural networks This paper describes a new method for setting the learning rate, named cyclical learning rates, which practically eliminates the need to experimentally find the best values and schedule for the global learning rates Instead of monotonically decreasing the learning rate, this method lets the learning rate cyclically vary between reasonable boundary values Training with cyclical learning rates instead of fixed values achieves improved classification accuracy without a need to tune and often in fewer iterations This paper also describes a simple way to estimate "reasonable bounds" -- linearly increasing the learning rate of the network for a few epochs In addition, cyclical learning rates are demonstrated on the CIFAR-10 and CIFAR-100 datasets with ResNets, Stochastic Depth networks, and DenseNets, and the ImageNet dataset with the AlexNet and GoogLeNet architectures These are practical tools for everyone who trains neural networks

Abstract: As computers become more pervasive in society, facilitating natural human---computer interaction (HCI) will have a positive impact on their use. Hence, there has been growing interest in the development of new approaches and technologies for bridging the human---computer barrier. The ultimate aim is to bring HCI to a regime where interactions with computers will be as natural as an interaction between humans, and to this end, incorporating gestures in HCI is an important research area. Gestures have long been considered as an interaction technique that can potentially deliver more natural, creative and intuitive methods for communicating with our computers. This paper provides an analysis of comparative surveys done in this area. The use of hand gestures as a natural interface serves as a motivating force for research in gesture taxonomies, its representations and recognition techniques, software platforms and frameworks which is discussed briefly in this paper. It focuses on the three main phases of hand gesture recognition i.e. detection, tracking and recognition. Different application which employs hand gestures for efficient interaction has been discussed under core and advanced application domains. This paper also provides an analysis of existing literature related to gesture recognition systems for human computer interaction by categorizing it under different key parameters. It further discusses the advances that are needed to further improvise the present hand gesture recognition systems for future perspective that can be widely used for efficient human computer interaction. The main goal of this survey is to provide researchers in the field of gesture based HCI with a summary of progress achieved to date and to help identify areas where further research is needed.

Abstract: Graph Convolutional Networks (GCNs) and their variants have experienced significant attention and have become the de facto methods for learning graph representations. GCNs derive inspiration primarily from recent deep learning approaches, and as a result, may inherit unnecessary complexity and redundant computation. In this paper, we reduce this excess complexity through successively removing nonlinearities and collapsing weight matrices between consecutive layers. We theoretically analyze the resulting linear model and show that it corresponds to a fixed low-pass filter followed by a linear classifier. Notably, our experimental evaluation demonstrates that these simplifications do not negatively impact accuracy in many downstream applications. Moreover, the resulting model scales to larger datasets, is naturally interpretable, and yields up to two orders of magnitude speedup over FastGCN.

Abstract: Background The recurrence score based on the 21-gene breast cancer assay predicts chemotherapy benefit if it is high and a low risk of recurrence in the absence of chemotherapy if it is low; however, there is uncertainty about the benefit of chemotherapy for most patients, who have a midrange score. Methods We performed a prospective trial involving 10,273 women with hormone-receptor–positive, human epidermal growth factor receptor 2 (HER2)–negative, axillary node–negative breast cancer. Of the 9719 eligible patients with follow-up information, 6711 (69%) had a midrange recurrence score of 11 to 25 and were randomly assigned to receive either chemoendocrine therapy or endocrine therapy alone. The trial was designed to show noninferiority of endocrine therapy alone for invasive disease–free survival (defined as freedom from invasive disease recurrence, second primary cancer, or death). Results Endocrine therapy was noninferior to chemoendocrine therapy in the analysis of invasive disease–free surv...

Abstract: So-called Weyl points can be thought of as magnetic monopoles in momentum space. Researchers show that certain transition-metal monophosphides are characterized by Weyl points.

Abstract: Inspired by the recent success of methods that employ shape priors to achieve robust 3D reconstructions, we propose a novel recurrent neural network architecture that we call the 3D Recurrent Reconstruction Neural Network (3D-R2N2). The network learns a mapping from images of objects to their underlying 3D shapes from a large collection of synthetic data [13]. Our network takes in one or more images of an object instance from arbitrary viewpoints and outputs a reconstruction of the object in the form of a 3D occupancy grid. Unlike most of the previous works, our network does not require any image annotations or object class labels for training or testing. Our extensive experimental analysis shows that our reconstruction framework (i) outperforms the state-of-the-art methods for single view reconstruction, and (ii) enables the 3D reconstruction of objects in situations when traditional SFM/SLAM methods fail (because of lack of texture and/or wide baseline).

Abstract: This report summarizes the tutorial presented by the author at NIPS 2016 on generative adversarial networks (GANs). The tutorial describes: (1) Why generative modeling is a topic worth studying, (2) how generative models work, and how GANs compare to other generative models, (3) the details of how GANs work, (4) research frontiers in GANs, and (5) state-of-the-art image models that combine GANs with other methods. Finally, the tutorial contains three exercises for readers to complete, and the solutions to these exercises.

Abstract: BACKGROUND: The economic burden of depression in the United States-including major depressive disorder (MDD), bipolar disorder, and dysthymia-was estimated at $831 billion in 2000 We update these findings using recent data, focusing on MDD alone and accounting for comorbid physical and psychiatric disorders METHOD: Using national survey (DSM-IV criteria) and administrative claims data (ICD-9 codes), we estimate the incremental economic burden of individuals with MDD as well as the share of these costs attributable to MDD, with attention to any changes that occurred between 2005 and 2010 RESULTS: The incremental economic burden of individuals with MDD increased by 215% (from $1732 billion to $2105 billion, inflation-adjusted dollars) The composition of these costs remained stable, with approximately 45% attributable to direct costs, 5% to suicide-related costs, and 50% to workplace costs Only 38% of the total costs were due to MDD itself as opposed to comorbid conditions CONCLUSIONS: Comorbid conditions account for the largest portion of the growing economic burden of MDD Future research should analyze further these comorbidities as well as the relative importance of factors contributing to that growing burden These include population growth, increase in MDD prevalence, increase in treatment cost per individual with MDD, changes in employment and treatment rates, as well as changes in the composition and quality of MDD treatment services Language: en

Abstract: We previously reported on MetaBAT, an automated metagenome binning software tool to reconstruct single genomes from microbial communities for subsequent analyses of uncultivated microbial species. MetaBAT has become one of the most popular binning tools largely due to its computational efficiency and ease of use, especially in binning experiments with a large number of samples and a large assembly. MetaBAT requires users to choose parameters to fine-tune its sensitivity and specificity. If those parameters are not chosen properly, binning accuracy can suffer, especially on assemblies of poor quality. Here, we developed MetaBAT 2 to overcome this problem. MetaBAT 2 uses a new adaptive binning algorithm to eliminate manual parameter tuning. We also performed extensive software engineering optimization to increase both computational and memory efficiency. Comparing MetaBAT 2 to alternative software tools on over 100 real world metagenome assemblies shows superior accuracy and computing speed. Binning a typical metagenome assembly takes only a few minutes on a single commodity workstation. We therefore recommend the community adopts MetaBAT 2 for their metagenome binning experiments. MetaBAT 2 is open source software and available at https://bitbucket.org/berkeleylab/metabat.

Abstract: New @ESCardio Guidelines for the Diagnosis and Management of Acute #PulmonaryEmbolism developed in collaboration with @EuroRespSoc now available: #cardiotwitter @erspublicationshttp://bit.ly/2HnrJaj

Abstract: Since the publication of Standards for QUality Improvement Reporting Excellence (SQUIRE 1.0) guidelines in 2008, the science of the field has advanced considerably. In this manuscript, we describe the development of SQUIRE 2.0 and its key components. We undertook the revision between 2012 and 2015 using (1) semistructured interviews and focus groups to evaluate SQUIRE 1.0 plus feedback from an international steering group, (2) two face-to-face consensus meetings to develop interim drafts and (3) pilot testing with authors and a public comment period. SQUIRE 2.0 emphasises the reporting of three key components of systematic efforts to improve the quality, value and safety of healthcare: the use of formal and informal theory in planning, implementing and evaluating improvement work; the context in which the work is done and the study of the intervention(s). SQUIRE 2.0 is intended for reporting the range of methods used to improve healthcare, recognising that they can be complex and multidimensional. It provides common ground to share these discoveries in the scholarly literature (http://www.squire-statement.org).

Abstract: Recent deep learning based approaches have shown promising results for the challenging task of inpainting large missing regions in an image. These methods can generate visually plausible image structures and textures, but often create distorted structures or blurry textures inconsistent with surrounding areas. This is mainly due to ineffectiveness of convolutional neural networks in explicitly borrowing or copying information from distant spatial locations. On the other hand, traditional texture and patch synthesis approaches are particularly suitable when it needs to borrow textures from the surrounding regions. Motivated by these observations, we propose a new deep generative model-based approach which can not only synthesize novel image structures but also explicitly utilize surrounding image features as references during network training to make better predictions. The model is a feed-forward, fully convolutional neural network which can process images with multiple holes at arbitrary locations and with variable sizes during the test time. Experiments on multiple datasets including faces (CelebA, CelebA-HQ), textures (DTD) and natural images (ImageNet, Places2) demonstrate that our proposed approach generates higher-quality inpainting results than existing ones. Code, demo and models are available at: this https URL.

Abstract: We generated genome-wide data from 69 Europeans who lived between 8,000-3,000 years ago by enriching ancient DNA libraries for a target set of almost 400,000 polymorphisms. Enrichment of these positions decreases the sequencing required for genome-wide ancient DNA analysis by a median of around 250-fold, allowing us to study an order of magnitude more individuals than previous studies and to obtain new insights about the past. We show that the populations of Western and Far Eastern Europe followed opposite trajectories between 8,000-5,000 years ago. At the beginning of the Neolithic period in Europe, ∼8,000-7,000 years ago, closely related groups of early farmers appeared in Germany, Hungary and Spain, different from indigenous hunter-gatherers, whereas Russia was inhabited by a distinctive population of hunter-gatherers with high affinity to a ∼24,000-year-old Siberian. By ∼6,000-5,000 years ago, farmers throughout much of Europe had more hunter-gatherer ancestry than their predecessors, but in Russia, the Yamnaya steppe herders of this time were descended not only from the preceding eastern European hunter-gatherers, but also from a population of Near Eastern ancestry. Western and Eastern Europe came into contact ∼4,500 years ago, as the Late Neolithic Corded Ware people from Germany traced ∼75% of their ancestry to the Yamnaya, documenting a massive migration into the heartland of Europe from its eastern periphery. This steppe ancestry persisted in all sampled central Europeans until at least ∼3,000 years ago, and is ubiquitous in present-day Europeans. These results provide support for a steppe origin of at least some of the Indo-European languages of Europe.