Abstract: Learning generic and robust feature representations with data from multiple domains for the same problem is of great value, especially for the problems that have multiple datasets but none of them are large enough to provide abundant data variations. In this work, we present a pipeline for learning deep feature representations from multiple domains with Convolutional Neural Networks (CNNs). When training a CNN with data from all the domains, some neurons learn representations shared across several domains, while some others are effective only for a specific one. Based on this important observation, we propose a Domain Guided Dropout algorithm to improve the feature learning procedure. Experiments show the effectiveness of our pipeline and the proposed algorithm. Our methods on the person re-identification problem outperform state-of-the-art methods on multiple datasets by large margins.

Abstract: Multi-gene and genomic datasets have become commonplace in the field of phylogenetics, but many of the existing tools are not designed for such datasets, which makes the analysis time-consuming and tedious. We therefore present PhyloSuite, a user-friendly workflow desktop platform dedicated to streamlining molecular sequence data management and evolutionary phylogenetics studies. It employs a plugin-based system that integrates a number of useful phylogenetic and bioinformatic tools, thereby streamlining the entire procedure, from data acquisition to phylogenetic tree annotation, with the following features: (i) point-and-click and drag-and-drop graphical user interface, (ii) a workspace to manage and organize molecular sequence data and results of analyses, (iii) GenBank entries extraction and comparative statistics, (iv) a phylogenetic workflow with batch processing capability, (v) elaborate bioinformatic analysis for mitochondrial genomes. The aim of PhyloSuite is to enable researchers to spend more time playing with scientific questions, instead of wasting it on conducting standard analyses. The compiled binary of PhyloSuite is available under the GPL license at https://github.com/dongzhang0725/PhyloSuite/releases, implemented in Python and runs on Windows, Mac OSX and Linux.

Abstract: The preparation of highly efficient perovskite nanocrystal light-emitting diodes is shown. A new trimethylaluminum vapor-based crosslinking method to render the nanocrystal films insoluble is applied. The resulting near-complete nanocrystal film coverage, coupled with the natural confinement of injected charges within the perovskite crystals, facilitates electron-hole capture and give rise to a remarkable electroluminescence yield of 5.7%.

Abstract: With the development of deep representation learning, the domain of reinforcement learning (RL) has become a powerful learning framework now capable of learning complex policies in high dimensional environments. This review summarises deep reinforcement learning (DRL) algorithms and provides a taxonomy of automated driving tasks where (D)RL methods have been employed, while addressing key computational challenges in real world deployment of autonomous driving agents. It also delineates adjacent domains such as behavior cloning, imitation learning, inverse reinforcement learning that are related but are not classical RL algorithms. The role of simulators in training agents, methods to validate, test and robustify existing solutions in RL are discussed.

Abstract: Machine learning offers a fantastically powerful toolkit for building useful complex prediction systems quickly. This paper argues it is dangerous to think of these quick wins as coming for free. Using the software engineering framework of technical debt, we find it is common to incur massive ongoing maintenance costs in real-world ML systems. We explore several ML-specific risk factors to account for in system design. These include boundary erosion, entanglement, hidden feedback loops, undeclared consumers, data dependencies, configuration issues, changes in the external world, and a variety of system-level anti-patterns.

Abstract: The purpose of this paper is to provide a typology of sampling designs for qualitative researchers. We introduce the following sampling strategies: (a) parallel sampling designs, which represent a body of sampling strategies that facilitate credible comparisons of two or more different subgroups that are extracted from the same levels of study; (b) nested sampling designs, which are sampling strategies that facilitate credible comparisons of two or more members of the same subgroup, wherein one or more members of the subgroup represent a sub-sample of the full sample; and (c) multilevel sampling designs, which represent sampling strategies that facilitate credible comparisons of two or more subgroups that are extracted from different levels of study. Key Words: Qualitative Research, Sampling Designs, Random Sampling, Purposive Sampling, and Sample Size

Abstract: This paper reports results from the randomized evaluation of a group-lending microcredit program in Hyderabad, India. A lender worked in 52 randomly selected neighborhoods, leading to an 8.4 percentage point increase in takeup of microcredit. Small business investment and profits of preexisting businesses increased, but consumption did not significantly increase. Durable goods expenditure increased, while “temptation goods” expenditure declined. We found no significant changes in health, education, or women’s empowerment. Two years later, after control areas had gained access to microcredit but households in treatment area had borrowed for longer and in larger amounts, very few significant differences persist. (JEL G21, G31, O16, O12, L25, I38)

Abstract: The Feedback In Realistic Environments (FIRE) project explores feedback in cosmological galaxy formation simulations. Previous FIRE simulations used an identical source code (“FIRE-1”) for consistency. Motivated by the development of more accurate numerics – including hydrodynamic solvers, gravitational softening, and supernova coupling algorithms – and exploration of new physics (e.g. magnetic fields), we introduce “FIRE-2”, an updated numerical implementation of FIRE physics for the GIZMO code. We run a suite of simulations and compare against FIRE-1: overall, FIRE-2 improvements do not qualitatively change galaxy-scale properties. We pursue an extensive study of numerics versus physics. Details of the star-formation algorithm, cooling physics, and chemistry have weak effects, provided that we include metal-line cooling and star formation occurs at higher-than-mean densities. We present new resolution criteria for high-resolution galaxy simulations. Most galaxy-scale properties are robust to numerics we test, provided: (1) Toomre masses are resolved; (2) feedback coupling ensures conservation, and (3) individual supernovae are time-resolved. Stellar masses and profiles are most robust to resolution, followed by metal abundances and morphologies, followed by properties of winds and circum-galactic media (CGM). Central (∼kpc) mass concentrations in massive (>L*) galaxies are sensitive to numerics (via trapping/recycling of winds in hot halos). Multiple feedback mechanisms play key roles: supernovae regulate stellar masses/winds; stellar mass-loss fuels late star formation; radiative feedback suppresses accretion onto dwarfs and instantaneous star formation in disks. We provide all initial conditions and numerical algorithms used.

Abstract: Background Coronavirus disease 2019 (COVID-19) can manifest as a viral-induced hyperinflammation with multiorgan involvement. Such patients often experience rapid deterioration and need for mechanical ventilation. Currently, no prospectively validated biomarker of impending respiratory failure is available. Objective We aimed to identify and prospectively validate biomarkers that allow the identification of patients in need of impending mechanical ventilation. Methods Patients with COVID-19 who were hospitalized from February 29 to April 9, 2020, were analyzed for baseline clinical and laboratory findings at admission and during the disease. Data from 89 evaluable patients were available for the purpose of analysis comprising an initial evaluation cohort (n = 40) followed by a temporally separated validation cohort (n = 49). Results We identified markers of inflammation, lactate dehydrogenase, and creatinine as the variables most predictive of respiratory failure in the evaluation cohort. Maximal IL-6 level before intubation showed the strongest association with the need for mechanical ventilation, followed by maximal CRP level. The respective AUC values for IL-6 and CRP levels in the evaluation cohort were 0.97 and 0.86, and they were similar in the validation cohort (0.90 and 0.83, respectively). The calculated optimal cutoff values during the course of disease from the evaluation cohort (IL-6 level > 80 pg/mL and CRP level > 97 mg/L) both correctly classified 80% of patients in the validation cohort regarding their risk of respiratory failure. Conclusion The maximal level of IL-6, followed by CRP level, was highly predictive of the need for mechanical ventilation. This suggests the possibility of using IL-6 or CRP level to guide escalation of treatment in patients with COVID-19–related hyperinflammatory syndrome.

Abstract: Firms that initiate dividend payments have positive earnings changes both before and after the dividend policy change, while those omitting dividend payments have negative earnings changes. Subsequent earnings changes are positively related to the dividend announcement return, and stock price reactions at subsequent earnings announcements are smaller than usual, suggesting that these earnings changes are partially anticipated at the dividend announcement. The results indicate that investors interpret announcements of dividend initiations and omissions as managers' forecasts of future earnings changes.

Abstract: PDBsum is a web server providing structural information on the entries in the Protein Data Bank (PDB). The analyses are primarily image-based and include protein secondary structure, protein-ligand and protein-DNA interactions, PROCHECK analyses of structural quality, and many others. The 3D structures can be viewed interactively in RasMol, PyMOL, and a JavaScript viewer called 3Dmol.js. Users can upload their own PDB files and obtain a set of password-protected PDBsum analyses for each. The server is freely accessible to all at: http://www.ebi.ac.uk/pdbsum.

Abstract: Judging others’ personalities is an essential skill in successful social living, as personality is a key driver behind people’s interactions, behaviors, and emotions. Although accurate personality judgments stem from social-cognitive skills, developments in machine learning show that computer models can also make valid judgments. This study compares the accuracy of human and computer-based personality judgments, using a sample of 86,220 volunteers who completed a 100-item personality questionnaire. We show that (i) computer predictions based on a generic digital footprint (Facebook Likes) are more accurate (r = 0.56) than those made by the participants’ Facebook friends using a personality questionnaire (r = 0.49); (ii) computer models show higher interjudge agreement; and (iii) computer personality judgments have higher external validity when predicting life outcomes such as substance use, political attitudes, and physical health; for some outcomes, they even outperform the self-rated personality scores. Computers outpacing humans in personality judgment presents significant opportunities and challenges in the areas of psychological assessment, marketing, and privacy.

Abstract: The global burden of disease (GBD) studies have derived detailed and comparable epidemiological and burden of disease estimates for schizophrenia. We report GBD 2016 estimates of schizophrenia prevalence and burden of disease with disaggregation by age, sex, year, and for all countries. We conducted a systematic review to identify studies reporting the prevalence, incidence, remission, and/or excess mortality associated with schizophrenia. Reported estimates which met our inclusion criteria were entered into a Bayesian meta-regression tool used in GBD 2016 to derive prevalence for 20 age groups, 7 super-regions, 21 regions, and 195 countries and territories. Burden of disease estimates were derived for acute and residual states of schizophrenia by multiplying the age-, sex-, year-, and location-specific prevalence by 2 disability weights representative of the disability experienced during these states. The systematic review found a total of 129 individual data sources. The global age-standardized point prevalence of schizophrenia in 2016 was estimated to be 0.28% (95% uncertainty interval [UI]: 0.24-0.31). No sex differences were observed in prevalence. Age-standardized point prevalence rates did not vary widely across countries or regions. Globally, prevalent cases rose from 13.1 (95% UI: 11.6-14.8) million in 1990 to 20.9 (95% UI: 18.5-23.4) million cases in 2016. Schizophrenia contributes 13.4 (95% UI: 9.9-16.7) million years of life lived with disability to burden of disease globally. Although schizophrenia is a low prevalence disorder, the burden of disease is substantial. Our modeling suggests that significant population growth and aging has led to a large and increasing disease burden attributable to schizophrenia, particularly for middle income countries.

Abstract: We discuss several classes of improvements to gravity solutions from the Gravity Recovery and Climate Experiment (GRACE) mission. These include both improvements in background geophysical models and orbital parameterization leading to the unconstrained spherical harmonic solution JPL RL05, and an alternate JPL RL05M mass concentration (mascon) solution benefitting from those same improvements but derived in surface spherical cap mascons. The mascon basis functions allow for convenient application of a priori information derived from near-global geophysical models to prevent striping in the solutions. The resulting mass flux solutions are shown to suffer less from leakage errors than harmonic solutions, and do not necessitate empirical filters to remove north-south stripes, lowering the dependence on using scale factors (the global mean scale factor decreases by 0.17) to gain accurate mass estimates. Ocean bottom pressure (OBP) time series derived from the mascon solutions are shown to have greater correlation with in situ data than do spherical harmonic solutions (increase in correlation coefficient of 0.08 globally), particularly in low-latitude regions with small signal power (increase in correlation coefficient of 0.35 regionally), in addition to reducing the error RMS with respect to the in situ data (reduction of 0.37 cm globally, and as much as 1 cm regionally). Greenland and Antarctica mass balance estimates derived from the mascon solutions agree within formal uncertainties with previously published results. Computing basin averages for hydrology applications shows general agreement between harmonic and mascon solutions for large basins; however, mascon solutions typically have greater resolution for smaller spatial regions, in particular when studying secular signals.

Abstract: CD8 + T cell–dependent killing of cancer cells requires efficient presentation of tumor antigens by human leukocyte antigen class I (HLA-I) molecules. However, the extent to which patient-specific HLA-I genotype influences response to anti–programmed cell death protein 1 or anti–cytotoxic T lymphocyte–associated protein 4 is currently unknown. We determined the HLA-I genotype of 1535 advanced cancer patients treated with immune checkpoint blockade (ICB). Maximal heterozygosity at HLA-I loci (“A,” “B,” and “C”) improved overall survival after ICB compared with patients who were homozygous for at least one HLA locus. In two independent melanoma cohorts, patients with the HLA-B44 supertype had extended survival, whereas the HLA-B62 supertype (including HLA-B*15:01) or somatic loss of heterozygosity at HLA-I was associated with poor outcome. Molecular dynamics simulations of HLA-B*15:01 revealed different elements that may impair CD8 + T cell recognition of neoantigens. Our results have important implications for predicting response to ICB and for the design of neoantigen-based therapeutic vaccines.

Abstract: Digital Twin technology is an emerging concept that has become the centre of attention for industry and, in more recent years, academia. The advancements in industry 4.0 concepts have facilitated its growth, particularly in the manufacturing industry. The Digital Twin is defined extensively but is best described as the effortless integration of data between a physical and virtual machine in either direction. The challenges, applications, and enabling technologies for Artificial Intelligence, Internet of Things (IoT) and Digital Twins are presented. A review of publications relating to Digital Twins is performed, producing a categorical review of recent papers. The review has categorised them by research areas: manufacturing, healthcare and smart cities, discussing a range of papers that reflect these areas and the current state of research. The paper provides an assessment of the enabling technologies, challenges and open research for Digital Twins.

Abstract: Trap-assisted recombination, despite being lower as compared with traditional inorganic solar cells, is still the dominant recombination mechanism in perovskite solar cells (PSCs) and limits their efficiency. We investigate the attributes of the primary trap-assisted recombination channels (grain boundaries and interfaces) and their correlation to defect ions in PSCs. We achieve this by using a validated device model to fit the simulations to the experimental data of efficient vacuum-deposited p–i–n and n–i–p CH3NH3PbI3 solar cells, including the light intensity dependence of the open-circuit voltage and fill factor. We find that, despite the presence of traps at interfaces and grain boundaries (GBs), their neutral (when filled with photogenerated charges) disposition along with the long-lived nature of holes leads to the high performance of PSCs. The sign of the traps (when filled) is of little importance in efficient solar cells with compact morphologies (fused GBs, low trap density). On the other hand,...

Abstract: This study examines the effect of shareholder proposals related to corporate social responsibility CSR on financial performance. Specifically, I focus on CSR proposals that pass or fail by a small margin of votes. The passage of such "close call" proposals is akin to a random assignment of CSR to companies and hence provides a quasi-experiment to study the effect of CSR on performance. I find that the adoption of close call CSR proposals leads to positive announcement returns and superior accounting performance, implying that these proposals are value enhancing. When I examine the channels through which companies benefit from CSR, I find that labor productivity and sales growth increase after the vote. Finally, I document that close call CSR proposals differ from non-close proposals along several dimensions. Accordingly, although my results imply that adopting close call CSR proposals is beneficial to companies, they do not necessarily imply that CSR proposals are beneficial in general. Data, as supplemental material, are available at http://dx.doi.org/10.1287/mnsc.2014.2038 . This paper was accepted by Wei Jiang, finance.

Abstract: The International Symposium on Biomedical Imaging (ISBI) held a grand challenge to evaluate computational systems for the automated detection of metastatic breast cancer in whole slide images of sentinel lymph node biopsies. Our team won both competitions in the grand challenge, obtaining an area under the receiver operating curve (AUC) of 0.925 for the task of whole slide image classification and a score of 0.7051 for the tumor localization task. A pathologist independently reviewed the same images, obtaining a whole slide image classification AUC of 0.966 and a tumor localization score of 0.733. Combining our deep learning system's predictions with the human pathologist's diagnoses increased the pathologist's AUC to 0.995, representing an approximately 85 percent reduction in human error rate. These results demonstrate the power of using deep learning to produce significant improvements in the accuracy of pathological diagnoses.

Abstract: We propose CornerNet, a new approach to object detection where we detect an object bounding box as a pair of keypoints, the top-left corner and the bottom-right corner, using a single convolution neural network. By detecting objects as paired keypoints, we eliminate the need for designing a set of anchor boxes commonly used in prior single-stage detectors. In addition to our novel formulation, we introduce corner pooling, a new type of pooling layer that helps the network better localize corners. Experiments show that CornerNet achieves a 42.2% AP on MS COCO, outperforming all existing one-stage detectors.

Abstract: We consider an agent's uncertainty about its environment and the problem of generalizing this uncertainty across states. Specifically, we focus on the problem of exploration in non-tabular reinforcement learning. Drawing inspiration from the intrinsic motivation literature, we use density models to measure uncertainty, and propose a novel algorithm for deriving a pseudo-count from an arbitrary density model. This technique enables us to generalize count-based exploration algorithms to the non-tabular case. We apply our ideas to Atari 2600 games, providing sensible pseudo-counts from raw pixels. We transform these pseudo-counts into exploration bonuses and obtain significantly improved exploration in a number of hard games, including the infamously difficult MONTEZUMA'S REVENGE.

Abstract: Advances in high-resolution cryo-electron microscopy (cryo-EM) require the development of validation metrics to independently assess map quality and model geometry. We report EMRinger, a tool that assesses the precise fitting of an atomic model into the map during refinement and shows how radiation damage alters scattering from negatively charged amino acids. EMRinger (https://github.com/fraser-lab/EMRinger) will be useful for monitoring progress in resolving and modeling high-resolution features in cryo-EM.

Abstract: Background Current phenotyping of chronic rhinosinusitis (CRS) into chronic rhinosinusitis with nasal polyps (CRSwNP) and chronic rhinosinusitis without nasal polyps (CRSsNP) might not adequately reflect the pathophysiologic diversity within patients with CRS. Objective We sought to identify inflammatory endotypes of CRS. Therefore we aimed to cluster patients with CRS based solely on immune markers in a phenotype-free approach. Secondarily, we aimed to match clusters to phenotypes. Methods In this multicenter case-control study patients with CRS and control subjects underwent surgery, and tissue was analyzed for IL-5, IFN-γ, IL-17A, TNF-α, IL-22, IL-1β, IL-6, IL-8, eosinophilic cationic protein, myeloperoxidase, TGF-β1, IgE, Staphylococcus aureus enterotoxin–specific IgE, and albumin. We used partition-based clustering. Results Clustering of 173 cases resulted in 10 clusters, of which 4 clusters with low or undetectable IL-5, eosinophilic cationic protein, IgE, and albumin concentrations, and 6 clusters with high concentrations of those markers. The group of IL-5–negative clusters, 3 clusters clinically resembled a predominant chronic rhinosinusitis without nasal polyps (CRSsNP) phenotype without increased asthma prevalence, and 1 cluster had a T H 17 profile and had mixed CRSsNP/CRSwNP. The IL-5–positive clusters were divided into a group with moderate IL-5 concentrations, a mixed CRSsNP/CRSwNP and increased asthma phenotype, and a group with high IL-5 levels, an almost exclusive nasal polyp phenotype with strongly increased asthma prevalence. In the latter group, 2 clusters demonstrated the highest concentrations of IgE and asthma prevalence, with all samples expressing Staphylococcus aureus enterotoxin–specific IgE. Conclusion Distinct CRS clusters with diverse inflammatory mechanisms largely correlated with phenotypes and further differentiated them and provided a more accurate description of the inflammatory mechanisms involved than phenotype information only.

Abstract: Molecular orientation in polymer solar cells is shown to play an important role in device performance.

Abstract: In light of the rising concerns about the spread of COVID-19 and calls to contain the Corona Virus, a growing number of tertiary institutions have shut down in regards to face-to-face classes globally. The Corona virus has revealed emerging vulnerabilities in education systems around the world. It is now clear that society needs flexible and resilient education systems as we face unpredictable futures. A meta-analysis methodology was adopted for this study and pertinent literature was visited to capture the essence of continued learning during these unprecedented times. Findings reveal that universities worldwide are moving more and more towards online learning or E- Learning. Findings also reveal that apart from resources, staff readiness, confidence, student accessibility and motivation play important function in ICT integrated learning. This exploratory paper proposes that staff members should use technology and technological gadgets to enhance learning especially during these exceptional times. Findings also propose online and remote learning as a necessity in times of lock downs and social distancing due to COVID-19 pandemic. It also provides a strong platform for further research.

Abstract: We introduce four new real-world distribution shift datasets consisting of changes in image style, image blurriness, geographic location, camera operation, and more With our new datasets, we take stock of previously proposed methods for improving out-of-distribution robustness and put them to the test We find that using larger models and artificial data augmentations can improve robustness on real-world distribution shifts, contrary to claims in prior work We find improvements in artificial robustness benchmarks can transfer to real-world distribution shifts, contrary to claims in prior work Motivated by our observation that data augmentations can help with real-world distribution shifts, we also introduce a new data augmentation method which advances the state-of-the-art and outperforms models pretrained with 1000 times more labeled data Overall we find that some methods consistently help with distribution shifts in texture and local image statistics, but these methods do not help with some other distribution shifts like geographic changes Our results show that future research must study multiple distribution shifts simultaneously, as we demonstrate that no evaluated method consistently improves robustness

Abstract: Background Convalescent plasma is frequently administered to patients with Covid-19 and has been reported, largely on the basis of observational data, to improve clinical outcomes. Minimal data are available from adequately powered randomized, controlled trials. Methods We randomly assigned hospitalized adult patients with severe Covid-19 pneumonia in a 2:1 ratio to receive convalescent plasma or placebo. The primary outcome was the patient's clinical status 30 days after the intervention, as measured on a six-point ordinal scale ranging from total recovery to death. Results A total of 228 patients were assigned to receive convalescent plasma and 105 to receive placebo. The median time from the onset of symptoms to enrollment in the trial was 8 days (interquartile range, 5 to 10), and hypoxemia was the most frequent severity criterion for enrollment. The infused convalescent plasma had a median titer of 1:3200 of total SARS-CoV-2 antibodies (interquartile range, 1:800 to 1:3200]. No patients were lost to follow-up. At day 30 day, no significant difference was noted between the convalescent plasma group and the placebo group in the distribution of clinical outcomes according to the ordinal scale (odds ratio, 0.83 (95% confidence interval [CI], 0.52 to 1.35; P = 0.46). Overall mortality was 10.96% in the convalescent plasma group and 11.43% in the placebo group, for a risk difference of -0.46 percentage points (95% CI, -7.8 to 6.8). Total SARS-CoV-2 antibody titers tended to be higher in the convalescent plasma group at day 2 after the intervention. Adverse events and serious adverse events were similar in the two groups. Conclusions No significant differences were observed in clinical status or overall mortality between patients treated with convalescent plasma and those who received placebo. (PlasmAr ClinicalTrials.gov number, NCT04383535.).