Abstract: Recent work has demonstrated that some functional categories of the genome contribute disproportionately to the heritability of complex diseases. Here we analyze a broad set of functional elements, including cell type-specific elements, to estimate their polygenic contributions to heritability in genome-wide association studies (GWAS) of 17 complex diseases and traits with an average sample size of 73,599. To enable this analysis, we introduce a new method, stratified LD score regression, for partitioning heritability from GWAS summary statistics while accounting for linked markers. This new method is computationally tractable at very large sample sizes and leverages genome-wide information. Our findings include a large enrichment of heritability in conserved regions across many traits, a very large immunological disease-specific enrichment of heritability in FANTOM5 enhancers and many cell type-specific enrichments, including significant enrichment of central nervous system cell types in the heritability of body mass index, age at menarche, educational attainment and smoking behavior.

Abstract: Background Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections and the resulting disease, coronavirus disease 2019 (Covid-19), have spread to millions of persons worldw...

Abstract: Robust detection and tracking of objects is crucial for the deployment of autonomous vehicle technology. Image based benchmark datasets have driven development in computer vision tasks such as object detection, tracking and segmentation of agents in the environment. Most autonomous vehicles, however, carry a combination of cameras and range sensors such as lidar and radar. As machine learning based methods for detection and tracking become more prevalent, there is a need to train and evaluate such methods on datasets containing range sensor data along with images. In this work we present nuTonomy scenes (nuScenes), the first dataset to carry the full autonomous vehicle sensor suite: 6 cameras, 5 radars and 1 lidar, all with full 360 degree field of view. nuScenes comprises 1000 scenes, each 20s long and fully annotated with 3D bounding boxes for 23 classes and 8 attributes. It has 7x as many annotations and 100x as many images as the pioneering KITTI dataset. We define novel 3D detection and tracking metrics. We also provide careful dataset analysis as well as baselines for lidar and image based detection and tracking. Data, development kit and more information are available online.

Abstract: BACKGROUND: Individuals often avoid or delay seeking professional help for mental health problems. Stigma may be a key deterrent to help-seeking but this has not been reviewed systematically. Our systematic review addressed the overarching question: What is the impact of mental health-related stigma on help-seeking for mental health problems? Subquestions were: (a) What is the size and direction of any association between stigma and help-seeking? (b) To what extent is stigma identified as a barrier to help-seeking? (c) What processes underlie the relationship between stigma and help-seeking? (d) Are there population groups for which stigma disproportionately deters help-seeking? METHOD: Five electronic databases were searched from 1980 to 2011 and references of reviews checked. A meta-synthesis of quantitative and qualitative studies, comprising three parallel narrative syntheses and subgroup analyses, was conducted. RESULTS: The review identified 144 studies with 90,189 participants meeting inclusion criteria. The median association between stigma and help-seeking was d = - 0.27, with internalized and treatment stigma being most often associated with reduced help-seeking. Stigma was the fourth highest ranked barrier to help-seeking, with disclosure concerns the most commonly reported stigma barrier. A detailed conceptual model was derived that describes the processes contributing to, and counteracting, the deterrent effect of stigma on help-seeking. Ethnic minorities, youth, men and those in military and health professions were disproportionately deterred by stigma. CONCLUSIONS: Stigma has a small- to moderate-sized negative effect on help-seeking. Review findings can be used to help inform the design of interventions to increase help-seeking.

Abstract: This chapter claims war makes states, and discusses that mercantile capitalism and state making reinforced each other The reflections that follow merely illustrate analogy of war making and state making with organized crime from few hundred years of European experience and offer tentative arguments concerning principles of change and variation underlying experience War making, extraction, and capital accumulation interacted to shape European state making Apologists for particular governments and for government in general commonly argue, precisely, that they offer protection from local and external violence Government's provision of protection, by this standard, often qualifies as racketeering European states built up their military apparatuses through sustained struggles with their subject populations and by means of selective extension of protection to different classes within those populations The agreements on protection constrained the rulers themselves, making them vulnerable to courts, to assemblies, to withdrawals of credit, services, and expertise

Abstract: Hydrogen technologies have experienced cycles of excessive expectations followed by disillusion. Nonetheless, a growing body of evidence suggests these technologies form an attractive option for the deep decarbonisation of global energy systems, and that recent improvements in their cost and performance point towards economic viability as well. This paper is a comprehensive review of the potential role that hydrogen could play in the provision of electricity, heat, industry, transport and energy storage in a low-carbon energy system, and an assessment of the status of hydrogen in being able to fulfil that potential. The picture that emerges is one of qualified promise: hydrogen is well established in certain niches such as forklift trucks, while mainstream applications are now forthcoming. Hydrogen vehicles are available commercially in several countries, and 225 000 fuel cell home heating systems have been sold. This represents a step change from the situation of only five years ago. This review shows that challenges around cost and performance remain, and considerable improvements are still required for hydrogen to become truly competitive. But such competitiveness in the medium-term future no longer seems an unrealistic prospect, which fully justifies the growing interest and policy support for these technologies around the world.

Abstract: The recent dramatic rise in power conversion efficiencies (PCEs) of perovskite solar cells (PSCs) has triggered intense research worldwide. However, high PCE values have often been reached with poor stability at an illuminated area of typically less than 0.1 square centimeter. We used heavily doped inorganic charge extraction layers in planar PSCs to achieve very rapid carrier extraction, even with 10- to 20-nanometer-thick layers, avoiding pinholes and eliminating local structural defects over large areas. The robust inorganic nature of the layers allowed for the fabrication of PSCs with an aperture area >1 square centimeter that have a PCE >15%, as certified by an accredited photovoltaic calibration laboratory. Hysteresis in the current-voltage characteristics was eliminated; the PSCs were stable, with >90% of the initial PCE remaining after 1000 hours of light soaking.

Abstract: A severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variant, VOC 202012/01 (lineage B.1.1.7), emerged in southeast England in September 2020 and is rapidly spreading toward fixation. Using a variety of statistical and dynamic modeling approaches, we estimate that this variant has a 43 to 90% (range of 95% credible intervals, 38 to 130%) higher reproduction number than preexisting variants. A fitted two-strain dynamic transmission model shows that VOC 202012/01 will lead to large resurgences of COVID-19 cases. Without stringent control measures, including limited closure of educational institutions and a greatly accelerated vaccine rollout, COVID-19 hospitalizations and deaths across England in the first 6 months of 2021 were projected to exceed those in 2020. VOC 202012/01 has spread globally and exhibits a similar transmission increase (59 to 74%) in Denmark, Switzerland, and the United States.

Abstract: Objective:To update and expand the 2013 Clinical Practice Guidelines for the Management of Pain, Agitation, and Delirium in Adult Patients in the ICU.Design:Thirty-two international experts, four methodologists, and four critical illness survivors met virtually at least monthly. All section groups g

Abstract: Microbes are dominant drivers of biogeochemical processes, yet drawing a global picture of functional diversity, microbial community structure, and their ecological determinants remains a grand challenge. We analyzed 7.2 terabases of metagenomic data from 243 Tara Oceans samples from 68 locations in epipelagic and mesopelagic waters across the globe to generate an ocean microbial reference gene catalog with >40 million nonredundant, mostly novel sequences from viruses, prokaryotes, and picoeukaryotes. Using 139 prokaryote-enriched samples, containing >35,000 species, we show vertical stratification with epipelagic community composition mostly driven by temperature rather than other environmental factors or geography. We identify ocean microbial core functionality and reveal that >73% of its abundance is shared with the human gut microbiome despite the physicochemical differences between these two ecosystems.

Abstract: Policy search methods can allow robots to learn control policies for a wide range of tasks, but practical applications of policy search often require hand-engineered components for perception, state estimation, and low-level control. In this paper, we aim to answer the following question: does training the perception and control systems jointly end-to-end provide better performance than training each component separately? To this end, we develop a method that can be used to learn policies that map raw image observations directly to torques at the robot's motors. The policies are represented by deep convolutional neural networks (CNNs) with 92,000 parameters, and are trained using a guided policy search method, which transforms policy search into supervised learning, with supervision provided by a simple trajectory-centric reinforcement learning method. We evaluate our method on a range of real-world manipulation tasks that require close coordination between vision and control, such as screwing a cap onto a bottle, and present simulated comparisons to a range of prior policy search methods.

Abstract: In recent years, several extreme weather disasters have partially or completely damaged regional crop production. While detailed regional accounts of the effects of extreme weather disasters exist, the global scale effects of droughts, floods and extreme temperature on crop production are yet to be quantified. Here we estimate for the first time, to our knowledge, national cereal production losses across the globe resulting from reported extreme weather disasters during 1964-2007. We show that droughts and extreme heat significantly reduced national cereal production by 9-10%, whereas our analysis could not identify an effect from floods and extreme cold in the national data. Analysing the underlying processes, we find that production losses due to droughts were associated with a reduction in both harvested area and yields, whereas extreme heat mainly decreased cereal yields. Furthermore, the results highlight ~7% greater production damage from more recent droughts and 8-11% more damage in developed countries than in developing ones. Our findings may help to guide agricultural priorities in international disaster risk reduction and adaptation efforts.

Abstract: Importance Previous analyses of obesity trends among children and adolescents showed an increase between 1988-1994 and 1999-2000, but no change between 2003-2004 and 2011-2012, except for a significant decline among children aged 2 to 5 years. Objectives To provide estimates of obesity and extreme obesity prevalence for children and adolescents for 2011-2014 and investigate trends by age between 1988-1994 and 2013-2014. Design, Setting, and Participants Children and adolescents aged 2 to 19 years with measured weight and height in the 1988-1994 through 2013-2014 National Health and Nutrition Examination Surveys. Exposures Survey period. Main Outcomes and Measures Obesity was defined as a body mass index (BMI) at or above the sex-specific 95th percentile on the US Centers for Disease Control and Prevention (CDC) BMI-for-age growth charts. Extreme obesity was defined as a BMI at or above 120% of the sex-specific 95th percentile on the CDC BMI-for-age growth charts. Detailed estimates are presented for 2011-2014. The analyses of linear and quadratic trends in prevalence were conducted using 9 survey periods. Trend analyses between 2005-2006 and 2013-2014 also were conducted. Results Measurements from 40 780 children and adolescents (mean age, 11.0 years; 48.8% female) between 1988-1994 and 2013-2014 were analyzed. Among children and adolescents aged 2 to 19 years, the prevalence of obesity in 2011-2014 was 17.0% (95% CI, 15.5%-18.6%) and extreme obesity was 5.8% (95% CI, 4.9%-6.8%). Among children aged 2 to 5 years, obesity increased from 7.2% (95% CI, 5.8%-8.8%) in 1988-1994 to 13.9% (95% CI, 10.7%-17.7%) ( P P = .03) in 2013-2014. Among children aged 6 to 11 years, obesity increased from 11.3% (95% CI, 9.4%-13.4%) in 1988-1994 to 19.6% (95% CI, 17.1%-22.4%) ( P P = .44). Obesity increased among adolescents aged 12 to 19 years between 1988-1994 (10.5% [95% CI, 8.8%-12.5%]) and 2013-2014 (20.6% [95% CI, 16.2%-25.6%]; P P = .02) and adolescents aged 12 to 19 years (2.6% [95% CI, 1.7%-3.9%] in 1988-1994 to 9.1% [95% CI, 7.0%-11.5%] in 2013-2014; P P value range, .09-.87). Conclusions and Relevance In this nationally representative study of US children and adolescents aged 2 to 19 years, the prevalence of obesity in 2011-2014 was 17.0% and extreme obesity was 5.8%. Between 1988-1994 and 2013-2014, the prevalence of obesity increased until 2003-2004 and then decreased in children aged 2 to 5 years, increased until 2007-2008 and then leveled off in children aged 6 to 11 years, and increased among adolescents aged 12 to 19 years.

Abstract: Accurate genome assembly is hampered by repetitive regions. Although long single molecule sequencing reads are better able to resolve genomic repeats than short-read data, most long-read assembly algorithms do not provide the repeat characterization necessary for producing optimal assemblies. Here, we present Flye, a long-read assembly algorithm that generates arbitrary paths in an unknown repeat graph, called disjointigs, and constructs an accurate repeat graph from these error-riddled disjointigs. We benchmark Flye against five state-of-the-art assemblers and show that it generates better or comparable assemblies, while being an order of magnitude faster. Flye nearly doubled the contiguity of the human genome assembly (as measured by the NGA50 assembly quality metric) compared with existing assemblers.

Abstract: Importance Despite the potential importance of understanding excess mortality among people with mental disorders, no comprehensive meta-analyses have been conducted quantifying mortality across mental disorders. Objective To conduct a systematic review and meta-analysis of mortality among people with mental disorders and examine differences in mortality risks by type of death, diagnosis, and study characteristics. Data sources We searched EMBASE, MEDLINE, PsychINFO, and Web of Science from inception through May 7, 2014, including references of eligible articles. Our search strategy included terms for mental disorders (eg, mental disorders, serious mental illness, and severe mental illness), specific diagnoses (eg, schizophrenia, depression, anxiety, and bipolar disorder), and mortality. We also used Google Scholar to identify articles that cited eligible articles. Study selection English-language cohort studies that reported a mortality estimate of mental disorders compared with a general population or controls from the same study setting without mental illness were included. Two reviewers independently reviewed the titles, abstracts, and articles. Of 2481 studies identified, 203 articles met the eligibility criteria and represented 29 countries in 6 continents. Data extraction and synthesis One reviewer conducted a full abstraction of all data, and 2 reviewers verified accuracy. Main outcomes and measures Mortality estimates (eg, standardized mortality ratios, relative risks, hazard ratios, odds ratios, and years of potential life lost) comparing people with mental disorders and the general population or people without mental disorders. We used random-effects meta-analysis models to pool mortality ratios for all, natural, and unnatural causes of death. We also examined years of potential life lost and estimated the population attributable risk of mortality due to mental disorders. Results For all-cause mortality, the pooled relative risk of mortality among those with mental disorders (from 148 studies) was 2.22 (95% CI, 2.12-2.33). Of these, 135 studies revealed that mortality was significantly higher among people with mental disorders than among the comparison population. A total of 67.3% of deaths among people with mental disorders were due to natural causes, 17.5% to unnatural causes, and the remainder to other or unknown causes. The median years of potential life lost was 10 years (n = 24 studies). We estimate that 14.3% of deaths worldwide, or approximately 8 million deaths each year, are attributable to mental disorders. Conclusions and relevance These estimates suggest that mental disorders rank among the most substantial causes of death worldwide. Efforts to quantify and address the global burden of illness need to better consider the role of mental disorders in preventable mortality.

Abstract: Batch Normalization (BN) is a milestone technique in the development of deep learning, enabling various networks to train. However, normalizing along the batch dimension introduces problems --- BN's error increases rapidly when the batch size becomes smaller, caused by inaccurate batch statistics estimation. This limits BN's usage for training larger models and transferring features to computer vision tasks including detection, segmentation, and video, which require small batches constrained by memory consumption. In this paper, we present Group Normalization (GN) as a simple alternative to BN. GN divides the channels into groups and computes within each group the mean and variance for normalization. GN's computation is independent of batch sizes, and its accuracy is stable in a wide range of batch sizes. On ResNet-50 trained in ImageNet, GN has 10.6% lower error than its BN counterpart when using a batch size of 2; when using typical batch sizes, GN is comparably good with BN and outperforms other normalization variants. Moreover, GN can be naturally transferred from pre-training to fine-tuning. GN can outperform its BN-based counterparts for object detection and segmentation in COCO, and for video classification in Kinetics, showing that GN can effectively replace the powerful BN in a variety of tasks. GN can be easily implemented by a few lines of code in modern libraries.

Abstract: We present an improved determination of the Hubble constant (H0) from Hubble Space Telescope (HST) observations of 70 long-period Cepheids in the Large Magellanic Cloud. These were obtained with the same WFC3 photometric system used to measure Cepheids in the hosts of Type Ia supernovae. Gyroscopic control of HST was employed to reduce overheads while collecting a large sample of widely-separated Cepheids. The Cepheid Period-Luminosity relation provides a zeropoint-free link with 0.4% precision between the new 1.2% geometric distance to the LMC from Detached Eclipsing Binaries (DEBs) measured by Pietrzynski et al (2019) and the luminosity of SNe Ia. Measurements and analysis of the LMC Cepheids were completed prior to knowledge of the new LMC distance. Combined with a refined calibration of the count-rate linearity of WFC3-IR with 0.1% precision (Riess et al 2019), these three improved elements together reduce the full uncertainty in the LMC geometric calibration of the Cepheid distance ladder from 2.5% to 1.3%. Using only the LMC DEBs to calibrate the ladder we find H0=74.22 +/- 1.82 km/s/Mpc including systematic uncertainties, 3% higher than before for this particular anchor. Combining the LMC DEBs, masers in NGC 4258 and Milky Way parallaxes yields our best estimate: H0 = 74.03 +/- 1.42 km/s/Mpc, including systematics, an uncertainty of 1.91%---15% lower than our best previous result. Removing any one of these anchors changes H0 by < 0.7%. The difference between H0 measured locally and the value inferred from Planck CMB+LCDM is 6.6+/-1.5 km/s/Mpc or 4.4 sigma (P=99.999% for Gaussian errors) in significance, raising the discrepancy beyond a plausible level of chance. We summarize independent tests which show this discrepancy is not readily attributable to an error in any one source or measurement, increasing the odds that it results from a cosmological feature beyond LambdaCDM.

Abstract: Several coronavirus disease 2019 (COVID-19) vaccines are currently in human trials. In June 2020, we surveyed 13,426 people in 19 countries to determine potential acceptance rates and factors influencing acceptance of a COVID-19 vaccine. Of these, 71.5% of participants reported that they would be very or somewhat likely to take a COVID-19 vaccine, and 48.1% reported that they would accept their employer's recommendation to do so. Differences in acceptance rates ranged from almost 90% (in China) to less than 55% (in Russia). Respondents reporting higher levels of trust in information from government sources were more likely to accept a vaccine and take their employer's advice to do so.

Abstract: This case series describes the clinical presentation, characteristics, and outcomes of patients with coronavirus disease 2019 (COVID-19) admitted to the intensive care unit at a public hospital in Washington State in February 2020, including initial reports of cardiomyopathy in one-third of the patients.

Abstract: Despite recent progress in understanding the cancer genome, there is still a relative delay in understanding the full aspects of the glycome and glycoproteome of cancer. Glycobiology has been instrumental in relevant discoveries in various biological and medical fields, and has contributed to the deciphering of several human diseases. Glycans are involved in fundamental molecular and cell biology processes occurring in cancer, such as cell signalling and communication, tumour cell dissociation and invasion, cell-matrix interactions, tumour angiogenesis, immune modulation and metastasis formation. The roles of glycans in cancer have been highlighted by the fact that alterations in glycosylation regulate the development and progression of cancer, serving as important biomarkers and providing a set of specific targets for therapeutic intervention. This Review discusses the role of glycans in fundamental mechanisms controlling cancer development and progression, and their applications in oncology.

Abstract: This paper proposes a novel model and dataset for 3D scene flow estimation with an application to autonomous driving. Taking advantage of the fact that outdoor scenes often decompose into a small number of independently moving objects, we represent each element in the scene by its rigid motion parameters and each superpixel by a 3D plane as well as an index to the corresponding object. This minimal representation increases robustness and leads to a discrete-continuous CRF where the data term decomposes into pairwise potentials between superpixels and objects. Moreover, our model intrinsically segments the scene into its constituting dynamic components. We demonstrate the performance of our model on existing benchmarks as well as a novel realistic dataset with scene flow ground truth. We obtain this dataset by annotating 400 dynamic scenes from the KITTI raw data collection using detailed 3D CAD models for all vehicles in motion. Our experiments also reveal novel challenges which cannot be handled by existing methods.

Abstract: Distributions of Earth’s species are changing at accelerating rates, increasingly driven by human-mediated climate change. Such changes are already altering the composition of ecological communities, but beyond conservation of natural systems, how and why does this matter? We review evidence that climate-driven species redistribution at regional to global scales affects ecosystem functioning, human well-being, and the dynamics of climate change itself. Production of natural resources required for food security, patterns of disease transmission, and processes of carbon sequestration are all altered by changes in species distribution. Consideration of these effects of biodiversity redistribution is critical yet lacking in most mitigation and adaptation strategies, including the United Nation’s Sustainable Development Goals.

Abstract: Background Patients with elevated triglyceride levels are at increased risk for ischemic events. Icosapent ethyl, a highly purified eicosapentaenoic acid ethyl ester, lowers triglyceride levels, but data are needed to determine its effects on ischemic events. Methods We performed a multicenter, randomized, double-blind, placebo-controlled trial involving patients with established cardiovascular disease or with diabetes and other risk factors, who had been receiving statin therapy and who had a fasting triglyceride level of 135 to 499 mg per deciliter (1.52 to 5.63 mmol per liter) and a low-density lipoprotein cholesterol level of 41 to 100 mg per deciliter (1.06 to 2.59 mmol per liter). The patients were randomly assigned to receive 2 g of icosapent ethyl twice daily (total daily dose, 4 g) or placebo. The primary end point was a composite of cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, coronary revascularization, or unstable angina. The key secondary end point was a composite of cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke. Results A total of 8179 patients were enrolled (70.7% for secondary prevention of cardiovascular events) and were followed for a median of 4.9 years. A primary end-point event occurred in 17.2% of the patients in the icosapent ethyl group, as compared with 22.0% of the patients in the placebo group (hazard ratio, 0.75; 95% confidence interval [CI], 0.68 to 0.83; P Conclusions Among patients with elevated triglyceride levels despite the use of statins, the risk of ischemic events, including cardiovascular death, was significantly lower among those who received 2 g of icosapent ethyl twice daily than among those who received placebo. (Funded by Amarin Pharma; REDUCE-IT ClinicalTrials.gov number, NCT01492361 .).

Abstract: This article is the American Cancer Society's biennial update on female breast cancer statistics in the United States, including data on incidence, mortality, survival, and screening. Over the most recent 5-year period (2012-2016), the breast cancer incidence rate increased slightly by 0.3% per year, largely because of rising rates of local stage and hormone receptor-positive disease. In contrast, the breast cancer death rate continues to decline, dropping 40% from 1989 to 2017 and translating to 375,900 breast cancer deaths averted. Notably, the pace of the decline has slowed from an annual decrease of 1.9% during 1998 through 2011 to 1.3% during 2011 through 2017, largely driven by the trend in white women. Consequently, the black-white disparity in breast cancer mortality has remained stable since 2011 after widening over the past 3 decades. Nevertheless, the death rate remains 40% higher in blacks (28.4 vs 20.3 deaths per 100,000) despite a lower incidence rate (126.7 vs 130.8); this disparity is magnified among black women aged <50 years, who have a death rate double that of whites. In the most recent 5-year period (2013-2017), the death rate declined in Hispanics (2.1% per year), blacks (1.5%), whites (1.0%), and Asians/Pacific Islanders (0.8%) but was stable in American Indians/Alaska Natives. However, by state, breast cancer mortality rates are no longer declining in Nebraska overall; in Colorado and Wisconsin in black women; and in Nebraska, Texas, and Virginia in white women. Breast cancer was the leading cause of cancer death in women (surpassing lung cancer) in four Southern and two Midwestern states among blacks and in Utah among whites during 2016-2017. Declines in breast cancer mortality could be accelerated by expanding access to high-quality prevention, early detection, and treatment services to all women.

Abstract: Interindividual clinical variability in the course of SARS-CoV-2 infection is immense. We report that at least 101 of 987 patients with life-threatening COVID-19 pneumonia had neutralizing IgG auto-Abs against IFN-ω (13 patients), the 13 types of IFN-α (36), or both (52), at the onset of critical disease; a few also had auto-Abs against the other three type I IFNs. The auto-Abs neutralize the ability of the corresponding type I IFNs to block SARS-CoV-2 infection in vitro. These auto-Abs were not found in 663 individuals with asymptomatic or mild SARS-CoV-2 infection and were present in only 4 of 1,227 healthy individuals. Patients with auto-Abs were aged 25 to 87 years and 95 were men. A B cell auto-immune phenocopy of inborn errors of type I IFN immunity underlies life-threatening COVID-19 pneumonia in at least 2.6% of women and 12.5% of men.

Abstract: Planar perovskite solar cells (PSCs) made entirely via solution processing at low temperatures (<150°C) offer promise for simple manufacturing, compatibility with flexible substrates, and perovskite-based tandem devices. However, these PSCs require an electron-selective layer that performs well with similar processing. We report a contact-passivation strategy using chlorine-capped TiO2 colloidal nanocrystal film that mitigates interfacial recombination and improves interface binding in low-temperature planar solar cells. We fabricated solar cells with certified efficiencies of 20.1 and 19.5% for active areas of 0.049 and 1.1 square centimeters, respectively, achieved via low-temperature solution processing. Solar cells with efficiency greater than 20% retained 90% (97% after dark recovery) of their initial performance after 500 hours of continuous room-temperature operation at their maximum power point under 1-sun illumination (where 1 sun is defined as the standard illumination at AM1.5, or 1 kilowatt/square meter).

Abstract: Deep neural networks are vulnerable to adversarial examples, which poses security concerns on these algorithms due to the potentially severe consequences. Adversarial attacks serve as an important surrogate to evaluate the robustness of deep learning models before they are deployed. However, most of existing adversarial attacks can only fool a black-box model with a low success rate. To address this issue, we propose a broad class of momentum-based iterative algorithms to boost adversarial attacks. By integrating the momentum term into the iterative process for attacks, our methods can stabilize update directions and escape from poor local maxima during the iterations, resulting in more transferable adversarial examples. To further improve the success rates for black-box attacks, we apply momentum iterative algorithms to an ensemble of models, and show that the adversarially trained models with a strong defense ability are also vulnerable to our black-box attacks. We hope that the proposed methods will serve as a benchmark for evaluating the robustness of various deep models and defense methods. With this method, we won the first places in NIPS 2017 Non-targeted Adversarial Attack and Targeted Adversarial Attack competitions.