Abstract: There is growing interest in automating neural network architecture design. Existing architecture search methods can be computationally expensive, requiring thousands of different architectures to be trained from scratch. Recent work has explored weight sharing across models to amortize the cost of training. Although previous methods reduced the cost of architecture search by orders of magnitude, they remain complex, requiring hypernetworks or reinforcement learning controllers. We aim to understand weight sharing for one-shot architecture search. With careful experimental analysis, we show that it is possible to efficiently identify promising architectures from a complex search space without either hypernetworks or RL.

Abstract: We capitalize on large amounts of unlabeled video in order to learn a model of scene dynamics for both video recognition tasks (e.g. action classification) and video generation tasks (e.g. future prediction). We propose a generative adversarial network for video with a spatio-temporal convolutional architecture that untangles the scene's foreground from the background. Experiments suggest this model can generate tiny videos up to a second at full frame rate better than simple baselines, and we show its utility at predicting plausible futures of static images. Moreover, experiments and visualizations show the model internally learns useful features for recognizing actions with minimal supervision, suggesting scene dynamics are a promising signal for representation learning. We believe generative video models can impact many applications in video understanding and simulation.

Abstract: Background: Previous studies have showed clinical characteristics of patients with the 2019 novel coronavirus disease (COVID-19) and the evidence of person-to-person transmission. Limited data are available for asymptomatic infections. This study aims to present the clinical characteristics of 24 cases with asymptomatic infection screened from close contacts and to show the transmission potential of asymptomatic COVID-19 virus carriers. Methods: Epidemiological investigations were conducted among all close contacts of COVID-19 patients (or suspected patients) in Nanjing, Jiangsu Province, China, from Jan 28 to Feb 9, 2020, both in clinic and in community. Asymptomatic carriers were laboratory-confirmed positive for the COVID-19 virus by testing the nucleic acid of the pharyngeal swab samples. Their clinical records, laboratory assessments, and chest CT scans were reviewed. Findings: None of the 24 asymptomatic cases presented any obvious symptoms before nucleic acid screening. Five cases (20.8%) developed symptoms (fever, cough, fatigue and etc.) during hospitalization. Twelve (50.0%) cases showed typical CT images of ground-glass chest and five (20.8%) presented stripe shadowing in the lungs. The remaining seven (29.2%) cases showed normal CT image and had no symptoms during hospitalization. These seven cases were younger (median age: 14.0 years; P = 0.012) than the rest. None of the 24 cases developed severe COVID-19 pneumonia or died. The median communicable period, defined as the interval from the first day of positive nucleic acid tests to the first day of continuous negative tests, was 9.5 days (up to 21 days among the 24 asymptomatic cases). Through epidemiological investigation, we observed a typical asymptomatic transmission to the cohabiting family members, which even caused severe COVID-19 pneumonia. Interpretation: The asymptomatic carriers identified from close contacts were prone to be mildly ill during hospitalization. However, the communicable period could be up to three weeks and the communicated patients could develop severe illness. These results highlighted the importance of close contact tracing and longitudinally surveillance via virus nucleic acid tests. Further isolation recommendation and continuous nucleic acid tests may also be recommended to the patients discharged.

Abstract: The low power conversion efficiency (PCE) of tin-based hybrid perovskite solar cells (HPSCs) is mainly attributed to the high background carrier density due to a high density of intrinsic defects such as Sn vacancies and oxidized species (Sn4+) that characterize Sn-based HPSCs. Herein, this study reports on the successful reduction of the background carrier density by more than one order of magnitude by depositing near-single-crystalline formamidinium tin iodide (FASnI3) films with the orthorhombic a-axis in the out-of-plane direction. Using these highly crystalline films, obtained by mixing a very small amount (0.08 m) of layered (2D) Sn perovskite with 0.92 m (3D) FASnI3, for the first time a PCE as high as 9.0% in a planar p–i–n device structure is achieved. These devices display negligible hysteresis and light soaking, as they benefit from very low trap-assisted recombination, low shunt losses, and more efficient charge collection. This represents a 50% improvement in PCE compared to the best reference cell based on a pure FASnI3 film using SnF2 as a reducing agent. Moreover, the 2D/3D-based HPSCs show considerable improved stability due to the enhanced robustness of the perovskite film compared to the reference cell.

Abstract: In this paper, we propose a novel unsupervised domain adaptation algorithm based on deep learning for visual object recognition. Specifically, we design a new model called Deep Reconstruction-Classification Network (DRCN), which jointly learns a shared encoding representation for two tasks: (i) supervised classification of labeled source data, and (ii) unsupervised reconstruction of unlabeled target data. In this way, the learnt representation not only preserves discriminability, but also encodes useful information from the target domain. Our new DRCN model can be optimized by using backpropagation similarly as the standard neural networks.

Abstract: Importance Patterns-of-failure studies suggest that in metastatic non–small-cell lung cancer (NSCLC) sites of gross disease at presentation are the first to progress when treated with chemotherapy. This knowledge has led to increased adoption of local ablative radiation therapy in patients with stage IV NSCLC, though prospective randomized evidence is limited. Objective To determine if intervening with noninvasive stereotactic ablative radiotherapy (SAbR) prior to maintenance chemotherapy in patients with non–progressive limited metastatic NSCLC after induction therapy led to significant improvements in progression-free survival (PFS). Design, Setting, and Participants This is a single-institution randomized phase 2 study of maintenance chemotherapy alone vs SAbR followed by maintenance chemotherapy for patients with limited metastatic NSCLC (primary plus up to 5 metastatic sites) whose tumors did not possessEGFR-targetable orALK-targetable mutations but did achieve a partial response or stable disease after induction chemotherapy. Interventions Maintenance chemotherapy or SAbR to all sites of gross disease (including SAbR or hypofractionated radiation to the primary) followed by maintenance chemotherapy. Main Outcomes and Measures The primary end point was PFS; secondary end points included toxic effects, local and distant tumor control, patterns of failure, and overall survival. Results A total of 29 patients (9 women and 20 men) were enrolled; 14 patients (median [range] age, 63.5 [51.0-78.0] years) were allocated to the SAbR-plus-maintenance chemotherapy arm, and 15 patients (median [range] age, 70.0 [51.0-79.0] years) were allocated to the maintenance chemotherapy–alone arm. The trial was stopped to accrual early after an interim analysis found a significant improvement in PFS in the SAbR-plus-maintenance chemotherapy arm of 9.7 months vs 3.5 months in the maintenance chemotherapy–alone arm (P = .01). Toxic effects were similar in both arms. There were no in-field failures with fewer overall recurrences in the SAbR arm while those patients receiving maintenance therapy alone had progression at existing sites of disease and distantly. Conclusions and Relevance Consolidative SAbR prior to maintenance chemotherapy appeared beneficial, nearly tripling PFS in patients with limited metastatic NSCLC compared with maintenance chemotherapy alone, with no difference in toxic effects. The irradiation prevented local failures in original disease, the most likely sites of first recurrence. Furthermore, PFS for patients with limited metastatic disease appeared similar to those patients with a greater metastatic burden, further arguing for the potential benefits of local therapy in limited metastatic settings. Trial Registration clinicaltrials.gov Identifier:NCT02045446

Abstract: We present HST photometry of a selected sample of 50 long-period, low-extinction Milky Way Cepheids measured on the same WFC3 F555W, F814W, and F160W-band photometric system as extragalactic Cepheids in SN Ia hosts. These bright Cepheids were observed with the WFC3 spatial scanning mode in the optical and near-infrared to mitigate saturation and reduce pixel-to-pixel calibration errors to reach a mean photometric error of 5 millimags per observation. We use the new Gaia DR2 parallaxes and HST photometry to simultaneously constrain the cosmic distance scale and to measure the DR2 parallax zeropoint offset appropriate for Cepheids. We find a value for the zeropoint offset of -46 +/- 13 muas or +/- 6 muas for a fixed distance scale, higher than found from quasars, as expected, for these brighter and redder sources. The precision of the distance scale from DR2 has been reduced by a factor of 2.5 due to the need to independently determine the parallax offset. The best fit distance scale is 1.006 +/- 0.033, relative to the scale from Riess et al 2016 with H0=73.24 km/s/Mpc used to predict the parallaxes photometrically, and is inconsistent with the scale needed to match the Planck 2016 CMB data combined with LCDM at the 2.9 sigma confidence level (99.6%). At 96.5% confidence we find that the formal DR2 errors may be underestimated as indicated. We identify additional error associated with the use of augmented Cepheid samples utilizing ground-based photometry and discuss their likely origins. Including the DR2 parallaxes with all prior distance ladder data raises the current tension between the late and early Universe route to the Hubble constant to 3.8 sigma (99.99 %). With the final expected precision from Gaia, the sample of 50 Cepheids with HST photometry will limit to 0.5% the contribution of the first rung of the distance ladder to the uncertainty in the Hubble constant.

Abstract: This paper provides a framework to analyze the implicit or explicit behavioral theories found in laws, regulations, and programs. The analysis focuses on policy tools or instruments and the underlying behavioral assumptions that guide their choice. We begin with the premise that public policy almost always attempts to get people to do things they otherwise would not have done, or it enables them to do things they might not have done otherwise. Policy tools are used to overcome impediments to policy-relevant actions. The five broad categories of tools we iden- tify-authority, incentives, capacity-building, symbolic and hortatory, and learning-make dif- ferent assumptions about how policy relevant behavior can be fostered. We contend that policy tools are essentially political phenomena, and that policy participation in the form of com- pliance, utilization, and other forms of "coproduction" is an important form of political behavior deserving of greater attention by political science.

Abstract: Social media comprises communication websites that facilitate relationship forming between users from diverse backgrounds, resulting in a rich social structure. User generated content encourages inquiry and decision-making. Given the relevance of social media to various stakeholders, it has received significant attention from researchers of various fields, including information systems. There exists no comprehensive review that integrates and synthesises the findings of literature on social media. This study discusses the findings of 132 papers (in selected IS journals) on social media and social networking published between 1997 and 2017. Most papers reviewed here examine the behavioural side of social media, investigate the aspect of reviews and recommendations, and study its integration for organizational purposes. Furthermore, many studies have investigated the viability of online communities/social media as a marketing medium, while others have explored various aspects of social media, including the risks associated with its use, the value that it creates, and the negative stigma attached to it within workplaces. The use of social media for information sharing during critical events as well as for seeking and/or rendering help has also been investigated in prior research. Other contexts include political and public administration, and the comparison between traditional and social media. Overall, our study identifies multiple emergent themes in the existing corpus, thereby furthering our understanding of advances in social media research. The integrated view of the extant literature that our study presents can help avoid duplication by future researchers, whilst offering fruitful lines of enquiry to help shape research for this emerging field.

Abstract: The scientific study of complex systems has transformed a wide range of disciplines in recent years, enabling researchers in both the natural and social sciences to model and predict phenomena as diverse as earthquakes, global warming, demographic patterns, financial crises, and the failure of materials. In this book, Didier Sornette boldly applies his varied experience in these areas to propose a simple, powerful, and general theory of how, why, and when stock markets crash. Most attempts to explain market failures seek to pinpoint triggering mechanisms that occur hours, days, or weeks before the collapse. Sornette proposes a radically different view: the underlying cause can be sought months and even years before the abrupt, catastrophic event in the build-up of cooperative speculation, which often translates into an accelerating rise of the market price, otherwise known as a "bubble." Anchoring his sophisticated, step-by-step analysis in leading-edge physical and statistical modeling techniques, he unearths remarkable insights and some predictions--among them, that the "end of the growth era" will occur around 2050. Sornette probes major historical precedents, from the decades-long "tulip mania" in the Netherlands that wilted suddenly in 1637 to the South Sea Bubble that ended with the first huge market crash in England in 1720, to the Great Crash of October 1929 and Black Monday in 1987, to cite just a few. He concludes that most explanations other than cooperative self-organization fail to account for the subtle bubbles by which the markets lay the groundwork for catastrophe. Any investor or investment professional who seeks a genuine understanding of looming financial disasters should read this book. Physicists, geologists, biologists, economists, and others will welcome Why Stock Markets Crash as a highly original "scientific tale," as Sornette aptly puts it, of the exciting and sometimes fearsome--but no longer quite so unfathomable--world of stock markets.

Abstract: Immune checkpoint inhibitors have shown significant therapeutic responses against tumors containing increased mutation-associated neoantigen load. We have examined the evolving landscape of tumor neoantigens during the emergence of acquired resistance in patients with non-small cell lung cancer after initial response to immune checkpoint blockade with anti-PD-1 or anti-PD-1/anti-CTLA-4 antibodies. Analyses of matched pretreatment and resistant tumors identified genomic changes resulting in loss of 7 to 18 putative mutation-associated neoantigens in resistant clones. Peptides generated from the eliminated neoantigens elicited clonal T-cell expansion in autologous T-cell cultures, suggesting that they generated functional immune responses. Neoantigen loss occurred through elimination of tumor subclones or through deletion of chromosomal regions containing truncal alterations, and was associated with changes in T-cell receptor clonality. These analyses provide insight into the dynamics of mutational landscapes during immune checkpoint blockade and have implications for the development of immune therapies that target tumor neoantigens.Significance: Acquired resistance to immune checkpoint therapy is being recognized more commonly. This work demonstrates for the first time that acquired resistance to immune checkpoint blockade can arise in association with the evolving landscape of mutations, some of which encode tumor neoantigens recognizable by T cells. These observations imply that widening the breadth of neoantigen reactivity may mitigate the development of acquired resistance. Cancer Discov; 7(3); 264-76. ©2017 AACR.See related commentary by Yang, p. 250This article is highlighted in the In This Issue feature, p. 235.

Abstract: Understanding varying levels of biodiversity within cities is pivotal to protect it in the face of global urbanisation. In the early stages of urban ecology studies on intra-urban biodiversity focused on the urban–rural gradient, representing a broad generalisation of features of the urban landscape. Increasingly, studies classify the urban landscape in more detail, quantifying separately the effects of individual urban features on biodiversity levels. However, while separate factors influencing biodiversity variation among cities worldwide have recently been analysed, a global analysis on the factors influencing biodiversity levels within cities is still lacking. We here present the first meta-analysis on intra-urban biodiversity variation across a large variety of taxonomic groups of 75 cities worldwide. Our results show that patch area and corridors have the strongest positive effects on biodiversity, complemented by vegetation structure. Local, biotic and management habitat variables were significantly more important than landscape, abiotic or design variables. Large sites greater than 50 ha are necessary to prevent a rapid loss of area-sensitive species. This indicates that, despite positive impacts of biodiversity-friendly management, increasing the area of habitat patches and creating a network of corridors is the most important strategy to maintain high levels of urban biodiversity.

Abstract: . Aerosol black carbon is a unique primary tracer for combustion emissions. It affects the optical properties of the atmosphere and is recognized as the second most important anthropogenic forcing agent for climate change. It is the primary tracer for adverse health effects caused by air pollution. For the accurate determination of mass equivalent black carbon concentrations in the air and for source apportionment of the concentrations, optical measurements by filter-based absorption photometers must take into account the "filter loading effect". We present a new real-time loading effect compensation algorithm based on a two parallel spot measurement of optical absorption. This algorithm has been incorporated into the new Aethalometer model AE33. Intercomparison studies show excellent reproducibility of the AE33 measurements and very good agreement with post-processed data obtained using earlier Aethalometer models and other filter-based absorption photometers. The real-time loading effect compensation algorithm provides the high-quality data necessary for real-time source apportionment and for determination of the temporal variation of the compensation parameter k.

Abstract: Importance The neuroinflammatory hypothesis of major depressive disorder is supported by several main findings. First, in humans and animals, activation of the immune system causes sickness behaviors that present during a major depressive episode (MDE), such as low mood, anhedonia, anorexia, and weight loss. Second, peripheral markers of inflammation are frequently reported in major depressive disorder. Third, neuroinflammatory illnesses are associated with high rates of MDEs. However, a fundamental limitation of the neuroinflammatory hypothesis is a paucity of evidence of brain inflammation during MDE. Translocator protein density measured by distribution volume (TSPO V T ) is increased in activated microglia, an important aspect of neuroinflammation. Objective To determine whether TSPO V T is elevated in the prefrontal cortex, anterior cingulate cortex (ACC), and insula in patients with MDE secondary to major depressive disorder. Design, Setting, and Participants Case-control study in a tertiary care psychiatric hospital from May 1, 2010, through February 1, 2014. Twenty patients with MDE secondary to major depressive disorder and 20 healthy control participants underwent positron emission tomography with fluorine F 18–labeled N -(2-(2-fluoroethoxy)benzyl)- N -(4-phenoxypyridin-3-yl)acetamide ([ 18 F]FEPPA). Patients with MDE were medication free for at least 6 weeks. All participants were otherwise healthy and nonsmokers. Main Outcomes and Measures Values of TSPO V T in the prefrontal cortex, ACC, and insula. Results In MDE, TSPO V T was significantly elevated in all brain regions examined (multivariate analysis of variance, F 15,23 = 4.5 [ P = .001]). The magnitude of TSPO V T elevation was 26% in the prefrontal cortex (mean [SD] TSPO V T , 12.5 [3.6] in patients with MDE and 10.0 [2.4] in controls), 32% in the ACC (mean [SD] TSPO V T , 12.3 [3.5] in patients with MDE and 9.3 [2.2] in controls), and 33% in the insula (mean [SD] TSPO V T , 12.9 [3.7] in patients with MDE and 9.7 [2.3] in controls). In MDE, greater TSPO V T in the ACC correlated with greater depression severity ( r = 0.63 [ P = .005]). Conclusions and Relevance This finding provides the most compelling evidence to date of brain inflammation, and more specifically microglial activation, in MDE. This finding is important for improving treatment because it implies that therapeutics that reduce microglial activation should be promising for MDE. The correlation between higher ACC TSPO V T and the severity of MDE is consistent with the concept that neuroinflammation in specific regions may contribute to sickness behaviors that overlap with the symptoms of MDE.

Abstract: Semantic scene understanding is important for various applications. In particular, self-driving cars need a fine-grained understanding of the surfaces and objects in their vicinity. Light detection and ranging (LiDAR) provides precise geometric information about the environment and is thus a part of the sensor suites of almost all self-driving cars. Despite the relevance of semantic scene understanding for this application, there is a lack of a large dataset for this task which is based on an automotive LiDAR. In this paper, we introduce a large dataset to propel research on laser-based semantic segmentation. We annotated all sequences of the KITTI Vision Odometry Benchmark and provide dense point-wise annotations for the complete 360-degree field-of-view of the employed automotive LiDAR. We propose three benchmark tasks based on this dataset: (i) semantic segmentation of point clouds using a single scan, (ii) semantic segmentation using multiple past scans, and (iii) semantic scene completion, which requires to anticipate the semantic scene in the future. We provide baseline experiments and show that there is a need for more sophisticated models to efficiently tackle these tasks. Our dataset opens the door for the development of more advanced methods, but also provides plentiful data to investigate new research directions.

Abstract: We present an analysis of the deepest Herschel images in four major extragalactic fields GOODS–North, GOODS–South, UDS, and COSMOS obtained within the GOODS–Herschel and CANDELS–Herschel key programs. The star formation picture provided by a total of 10 497 individual far-infrared detections is supplemented by the stacking analysis of a mass complete sample of 62 361 star-forming galaxies from the Hubble Space Telescope (HST) H band-selected catalogs of the CANDELS survey and from two deep ground-based Ks band-selected catalogs in the GOODS–North and the COSMOS-wide field to obtain one of the most accurate and unbiased understanding to date of the stellar mass growth over the cosmic history. We show, for the first time, that stacking also provides a powerful tool to determine the dispersion of a physical correlation and describe our method called “scatter stacking”, which may be easily generalized to other experiments. The combination of direct UV and far-infrared UV-reprocessed light provides a complete census on the star formation rates (SFRs), allowing us to demonstrate that galaxies at z = 4 to 0 of all stellar masses (M∗) follow a universal scaling law, the so-called main sequence of star-forming galaxies. We find a universal close-to-linear slope of the log 10(SFR)–log 10(M∗) relation, with evidence for a flattening of the main sequence at high masses (log 10(M∗/M⊙) > 10.5) that becomesless prominent with increasing redshift and almost vanishes by z ≃ 2. This flattening may be due to the parallel stellar growth of quiescent bulges in star-forming galaxies, which mostly happens over the same redshift range. Within the main sequence, we measure a nonvarying SFR dispersion of 0.3 dex: at a fixed redshift and stellar mass, about 68% of star-forming galaxies form stars at a universal rate within a factor 2. The specific SFR (sSFR = SFR/M∗) of star-forming galaxies is found to continuously increase from z = 0 to 4. Finally we discuss the implications of our findings on the cosmic SFR history and on the origin of present-day stars: more than two-thirds of present-day stars must have formed in a regime dominated by the “main sequence” mode. As a consequence we conclude that, although omnipresent in the distant Universe, galaxy mergers had little impact in shaping the global star formation history over the last 12.5 billion years.

Abstract: Programmed cell death–1 (PD-1)–targeted therapies enhance T cell responses and show efficacy in multiple cancers, but the role of costimulatory molecules in this T cell rescue remains elusive. Here, we demonstrate that the CD28/B7 costimulatory pathway is essential for effective PD-1 therapy during chronic viral infection. Conditional gene deletion showed a cell-intrinsic requirement of CD28 for CD8 T cell proliferation after PD-1 blockade. B7-costimulation was also necessary for effective PD-1 therapy in tumor-bearing mice. In addition, we found that CD8 T cells proliferating in blood after PD-1 therapy of lung cancer patients were predominantly CD28-positive. Taken together, these data demonstrate CD28-costimulation requirement for CD8 T cell rescue and suggest an important role for the CD28/B7 pathway in PD-1 therapy of cancer patients.

Abstract: Importance Current information on the prevalence and sociodemographic and clinical profiles of individuals in the general population with DSM-5 drug use disorder (DUD) is limited. Given the present societal and economic context in the United States and the new diagnostic system, up-to-date national information is needed from a single uniform data source. Objective To present nationally representative findings on the prevalence, correlates, psychiatric comorbidity, disability, and treatment of DSM-5 DUD diagnoses overall and by severity level. Design, Setting, and Participants In-person interviews were conducted with 36 309 adults in the 2012-2013 National Epidemiologic Survey on Alcohol and Related Conditions–III, a cross-sectional representative survey of the United States. The household response rate was 72%; person-level response rate, 84%; and overall response rate, 60.1%. Data were collected April 2012 through June 2013 and analyzed from February through March 2015. Main Outcomes and Measures Twelve-month and lifetime DUD, based on amphetamine, cannabis, club drug, cocaine, hallucinogen, heroin, nonheroin opioid, sedative/tranquilizer, and/or solvent/inhalant use disorders. Results Prevalences of 12-month and lifetime DUD were 3.9% and 9.9%, respectively. Drug use disorder was generally greater among men, white and Native American individuals, younger and previously or never married adults, those with lower education and income, and those residing in the West. Significant associations were found between 12-month and lifetime DUD and other substance use disorders. Significant associations were also found between any 12-month DUD and major depressive disorder (odds ratio [OR], 1.3; 95% CI, 1.09-1.64), dysthymia (OR, 1.5; 95% CI, 1.09-2.02), bipolar I (OR, 1.5; 95% CI, 1.06-2.05), posttraumatic stress disorder (OR, 1.6; 95% CI, 1.27-2.10), and antisocial (OR, 1.4; 95% CI, 1.11-1.75), borderline (OR, 1.8; 95% CI, 1.41-2.24), and schizotypal (OR, 1.5; 95% CI, 1.18-1.87) personality disorders. Similar associations were found for any lifetime DUD with the exception that lifetime DUD was also associated with generalized anxiety disorder (OR, 1.3; 95% CI, 1.06-1.49), panic disorder (OR, 1.3; 95% CI, 1.06-1.59), and social phobia (OR, 1.3; 95% CI, 1.09-1.64). Twelve-month DUD was associated with significant disability, increasing with DUD severity. Among respondents with 12-month and lifetime DUD, only 13.5% and 24.6% received treatment, respectively. Conclusions and Relevance DSM-5 DUD is a common, highly comorbid, and disabling disorder that largely goes untreated in the United States. These findings indicate the need for additional studies to understand the broad relationships in more detail; estimate present-day economic costs of DUDs; investigate hypotheses regarding etiology, chronicity, and treatment use; and provide information to policy makers about allocation of resources for service delivery and research. Findings also indicate an urgent need to destigmatize DUD and educate the public, clinicians, and policy makers about its treatment to encourage affected individuals to obtain help.

Abstract: The increasing amount of electronic word of mouth (eWOM) has significantly affected the way consumers make purchase decisions. Empirical studies have established an effect of eWOM on sales but disagree on which online platforms, products, and eWOM metrics moderate this effect. The authors conduct a meta-analysis of 1,532 effect sizes across 96 studies covering 40 platforms and 26 product categories. On average, eWOM is positively correlated with sales (.091), but its effectiveness differs across platform, product, and metric factors. For example, the effectiveness of eWOM on social media platforms is stronger when eWOM receivers can assess their own similarity to eWOM senders, whereas these homophily details do not influence the effectiveness of eWOM for e-commerce platforms. In addition, whereas eWOM has a stronger effect on sales for tangible goods new to the market, the product life cycle does not moderate the eWOM effectiveness for services. With respect to the eWOM metrics, eWOM volume has a ...

Abstract: Lithium–sulfur batteries hold great promise for serving as next generation high energy density batteries. However, the shuttle of polysulfide induces rapid capacity degradation and poor cycling stability of lithium–sulfur cells. Herein, we proposed a unique lithium–sulfur battery configuration with an ultrathin graphene oxide (GO) membrane for high stability. The oxygen electronegative atoms modified GO into a polar plane, and the carboxyl groups acted as ion-hopping sites of positively charged species (Li+) and rejected the transportation of negatively charged species (Sn2–) due to the electrostatic interactions. Such electrostatic repulsion and physical inhibition largely decreased the transference of polysulfides across the GO membrane in the lithium–sulfur system. Consequently, the GO membrane with highly tunable functionalization properties, high mechanical strength, low electric conductivity, and facile fabrication procedure is an effective permselective separator system in lithium–sulfur batteries....

Abstract: This paper shows that pretraining multilingual language models at scale leads to significant performance gains for a wide range of cross-lingual transfer tasks. We train a Transformer-based masked language model on one hundred languages, using more than two terabytes of filtered CommonCrawl data. Our model, dubbed XLM-R, significantly outperforms multilingual BERT (mBERT) on a variety of cross-lingual benchmarks, including +14.6% average accuracy on XNLI, +13% average F1 score on MLQA, and +2.4% F1 score on NER. XLM-R performs particularly well on low-resource languages, improving 15.7% in XNLI accuracy for Swahili and 11.4% for Urdu over previous XLM models. We also present a detailed empirical analysis of the key factors that are required to achieve these gains, including the trade-offs between (1) positive transfer and capacity dilution and (2) the performance of high and low resource languages at scale. Finally, we show, for the first time, the possibility of multilingual modeling without sacrificing per-language performance; XLM-R is very competitive with strong monolingual models on the GLUE and XNLI benchmarks. We will make our code, data and models publicly available.

Abstract: We present Chronos, a system that enables a single WiFi access point to localize clients to within tens of centimeters. Such a system can bring indoor positioning to homes and small businesses which typically have a single access point. The key enabler underlying Chronos is a novel algorithm that can compute sub-nanosecond time-of-flight using commodity WiFi cards. By multiplying the time-of-flight with the speed of light, a MIMO access point computes the distance between each of its antennas and the client, hence localizing it. Our implementation on commodity WiFi cards demonstrates that Chronos's accuracy is comparable to state-of-the-art localization systems, which use four or five access points.

Abstract: The Genotype-Tissue Expression (GTEx) project, sponsored by the NIH Common Fund, was established to study the correlation between human genetic variation and tissue-specific gene expression in non-diseased individuals. A significant challenge was the collection of high-quality biospecimens for extensive genomic analyses. Here we describe how a successful infrastructure for biospecimen procurement was developed and implemented by multiple research partners to support the prospective collection, annotation, and distribution of blood, tissues, and cell lines for the GTEx project. Other research projects can follow this model and form beneficial partnerships with rapid autopsy and organ procurement organizations to collect high quality biospecimens and associated clinical data for genomic studies. Biospecimens, clinical and genomic data, and Standard Operating Procedures guiding biospecimen collection for the GTEx project are available to the research community.

Abstract: Evidence from mouse chronic viral infection models suggests that CD8+ T cell subsets characterized by distinct expression levels of the receptor PD-1 diverge in their state of exhaustion and potential for reinvigoration by PD-1 blockade. However, it remains unknown whether T cells in human cancer adopt a similar spectrum of exhausted states based on PD-1 expression levels. We compared transcriptional, metabolic and functional signatures of intratumoral CD8+ T lymphocyte populations with high (PD-1T), intermediate (PD-1N) and no PD-1 expression (PD-1–) from non-small-cell lung cancer patients. PD-1T T cells showed a markedly different transcriptional and metabolic profile from PD-1N and PD-1– lymphocytes, as well as an intrinsically high capacity for tumor recognition. Furthermore, while PD-1T lymphocytes were impaired in classical effector cytokine production, they produced CXCL13, which mediates immune cell recruitment to tertiary lymphoid structures. Strikingly, the presence of PD-1T cells was strongly predictive for both response and survival in a small cohort of non-small-cell lung cancer patients treated with PD-1 blockade. The characterization of a distinct state of tumor-reactive, PD-1-bright lymphocytes in human cancer, which only partially resembles that seen in chronic infection, provides potential avenues for therapeutic intervention. Tumor-infiltrating CD8+ T cells with high expression of PD-1 in non-small-cell lung cancer are distinct from exhausted T cells in chronic virus infection, have high tumor reactivity and associate with response to PD-1-targeted immunotherapy.

Abstract: Functionalization of atomically thin nanomaterials enables the tailoring of their chemical, optical and electronic properties. Exfoliated black phosphorus (BP)-a layered two-dimensional semiconductor-exhibits favourable charge-carrier mobility, tunable bandgap and highly anisotropic properties, but it is chemically reactive and degrades rapidly in ambient conditions. Here we show that covalent aryl diazonium functionalization suppresses the chemical degradation of exfoliated BP even after three weeks of ambient exposure. This chemical modification scheme spontaneously forms phosphorus-carbon bonds, has a reaction rate sensitive to the aryl diazonium substituent and alters the electronic properties of exfoliated BP, ultimately yielding a strong, tunable p-type doping that simultaneously improves the field-effect transistor mobility and on/off current ratio. This chemical functionalization pathway controllably modifies the properties of exfoliated BP, and thus improves its prospects for nanoelectronic applications.

Abstract: In this paper, we introduce the Intervention Mapping (IM) taxonomy of behaviour change methods and its potential to be developed into a coding taxonomy. That is, although IM and its taxonomy of behaviour change methods are not in fact new, because IM was originally developed as a tool for intervention development, this potential was not immediately apparent. Second, in explaining the IM taxonomy and defining the relevant constructs, we call attention to the existence of parameters for effectiveness of methods, and explicate the related distinction between theory-based methods and practical applications and the probability that poor translation of methods may lead to erroneous conclusions as to method-effectiveness. Third, we recommend a minimal set of intervention characteristics that may be reported when intervention descriptions and evaluations are published. Specifying these characteristics can greatly enhance the quality of our meta-analyses and other literature syntheses. In conclusion, the dynamics of behaviour change are such that any taxonomy of methods of behaviour change needs to acknowledge the importance of, and provide instruments for dealing with, three conditions for effectiveness for behaviour change methods. For a behaviour change method to be effective: (1) it must target a determinant that predicts behaviour; (2) it must be able to change that determinant; (3) it must be translated into a practical application in a way that preserves the parameters for effectiveness and fits with the target population, culture, and context. Thus, taxonomies of methods of behaviour change must distinguish the specific determinants that are targeted, practical, specific applications, and the theory-based methods they embody. In addition, taxonomies should acknowledge that the lists of behaviour change methods will be used by, and should be used by, intervention developers. Ideally, the taxonomy should be readily usable for this goal; but alternatively, it should be clear how the information in the taxonomy can be used in practice. The IM taxonomy satisfies these requirements, and it would be beneficial if other taxonomies would be extended to also meet these needs.

Abstract: We develop a fully automatic image colorization system. Our approach leverages recent advances in deep networks, exploiting both low-level and semantic representations. As many scene elements naturally appear according to multimodal color distributions, we train our model to predict per-pixel color histograms. This intermediate output can be used to automatically generate a color image, or further manipulated prior to image formation. On both fully and partially automatic colorization tasks, we outperform existing methods. We also explore colorization as a vehicle for self-supervised visual representation learning.