Abstract: The severe acute respiratory syndrome coronavirus 2 (SARS-Cov-2), the pathogen of 2019 novel coronavirus disease (COVID-19), has posed a serious threat to global public health. The WHO has declared the outbreak of SARS-CoV-2 infection an international public health emergency. Lung lesions have been considered as the major damage caused by SARS-CoV-2 infection. However, liver injury has also been reported to occur during the course of the disease in severe cases. Similarly, previous studies have shown that liver damage was common in the patients infected by the other two highly pathogenic coronavirus - severe acute respiratory syndrome coronavirus (SARS-CoV) and the Middle East respiratory syndrome coronavirus (MERS-CoV), and associated with the severity of diseases. In this review, the characteristics and mechanism of liver injury caused by SARS-CoV, MERS-CoV as well as SARS-CoV-2 infection were summarized, which may provide help for further studies on the liver injury of COVID-19.

Abstract: The field of defense strategies against adversarial attacks has significantly grown over the last years, but progress is hampered as the evaluation of adversarial defenses is often insufficient and thus gives a wrong impression of robustness. Many promising defenses could be broken later on, making it difficult to identify the state-of-the-art. Frequent pitfalls in the evaluation are improper tuning of hyperparameters of the attacks, gradient obfuscation or masking. In this paper we first propose two extensions of the PGD-attack overcoming failures due to suboptimal step size and problems of the objective function. We then combine our novel attacks with two complementary existing ones to form a parameter-free, computationally affordable and user-independent ensemble of attacks to test adversarial robustness. We apply our ensemble to over 50 models from papers published at recent top machine learning and computer vision venues. In all except one of the cases we achieve lower robust test accuracy than reported in these papers, often by more than $10\%$, identifying several broken defenses.

Abstract: Effectively leveraging large, previously collected datasets in reinforcement learning (RL) is a key challenge for large-scale real-world applications. Offline RL algorithms promise to learn effective policies from previously-collected, static datasets without further interaction. However, in practice, offline RL presents a major challenge, and standard off-policy RL methods can fail due to overestimation of values induced by the distributional shift between the dataset and the learned policy, especially when training on complex and multi-modal data distributions. In this paper, we propose conservative Q-learning (CQL), which aims to address these limitations by learning a conservative Q-function such that the expected value of a policy under this Q-function lower-bounds its true value. We theoretically show that CQL produces a lower bound on the value of the current policy and that it can be incorporated into a policy learning procedure with theoretical improvement guarantees. In practice, CQL augments the standard Bellman error objective with a simple Q-value regularizer which is straightforward to implement on top of existing deep Q-learning and actor-critic implementations. On both discrete and continuous control domains, we show that CQL substantially outperforms existing offline RL methods, often learning policies that attain 2-5 times higher final return, especially when learning from complex and multi-modal data distributions.

Abstract: The Integrated Microbial Genomes & Microbiomes system v.5.0 (IMG/M: https://img.jgi.doe.gov/m/) contains annotated datasets categorized into: archaea, bacteria, eukarya, plasmids, viruses, genome fragments, metagenomes, cell enrichments, single particle sorts, and metatranscriptomes. Source datasets include those generated by the DOE's Joint Genome Institute (JGI), submitted by external scientists, or collected from public sequence data archives such as NCBI. All submissions are typically processed through the IMG annotation pipeline and then loaded into the IMG data warehouse. IMG's web user interface provides a variety of analytical and visualization tools for comparative analysis of isolate genomes and metagenomes in IMG. IMG/M allows open access to all public genomes in the IMG data warehouse, while its expert review (ER) system (IMG/MER: https://img.jgi.doe.gov/mer/) allows registered users to access their private genomes and to store their private datasets in workspace for sharing and for further analysis. IMG/M data content has grown by 60% since the last report published in the 2017 NAR Database Issue. IMG/M v.5.0 has a new and more powerful genome search feature, new statistical tools, and supports metagenome binning.

Abstract: In the recent past, the computer vision community has developed centralized benchmarks for the performance evaluation of a variety of tasks, including generic object and pedestrian detection, 3D reconstruction, optical flow, single-object short-term tracking, and stereo estimation. Despite potential pitfalls of such benchmarks, they have proved to be extremely helpful to advance the state of the art in the respective area. Interestingly, there has been rather limited work on the standardization of quantitative benchmarks for multiple target tracking. One of the few exceptions is the well-known PETS dataset, targeted primarily at surveillance applications. Despite being widely used, it is often applied inconsistently, for example involving using different subsets of the available data, different ways of training the models, or differing evaluation scripts. This paper describes our work toward a novel multiple object tracking benchmark aimed to address such issues. We discuss the challenges of creating such a framework, collecting existing and new data, gathering state-of-the-art methods to be tested on the datasets, and finally creating a unified evaluation system. With MOTChallenge we aim to pave the way toward a unified evaluation framework for a more meaningful quantification of multi-target tracking.

Abstract: We analyze chiral topological edge modes in a non-Hermitian variant of the 2D Dirac equation. Such modes appear at interfaces between media with different "masses" and/or signs of the "non-Hermitian charge." The existence of these edge modes is intimately related to exceptional points of the bulk Hamiltonians, i.e., degeneracies in the bulk spectra of the media. We find that the topological edge modes can be divided into three families ("Hermitian-like," "non-Hermitian," and "mixed"); these are characterized by two winding numbers, describing two distinct kinds of half-integer charges carried by the exceptional points. We show that all the above types of topological edge modes can be realized in honeycomb lattices of ring resonators with asymmetric or gain-loss couplings.

Abstract: Summary Background Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was reported from China in January, 2020. SARS-CoV-2 is efficiently transmitted from person to person and, in 2 months, has caused more than 82 000 laboratory-confirmed cases of coronavirus disease 2019 (COVID-19) and 2800 deaths in 46 countries. The total number of cases and deaths has surpassed that of the 2003 severe acute respiratory syndrome coronavirus (SARS-CoV). Although both COVID-19 and severe acute respiratory syndrome (SARS) manifest as pneumonia, COVID-19 is associated with apparently more efficient transmission, fewer cases of diarrhoea, increased mental confusion, and a lower crude fatality rate. However, the underlying virus–host interactive characteristics conferring these observations on transmissibility and clinical manifestations of COVID-19 remain unknown. Methods We systematically investigated the cellular susceptibility, species tropism, replication kinetics, and cell damage of SARS-CoV-2 and compared findings with those for SARS-CoV. We compared SARS-CoV-2 and SARS-CoV replication in different cell lines with one-way ANOVA. For the area under the curve comparison between SARS-CoV-2 and SARS-CoV replication in Calu3 (pulmonary) and Caco2 (intestinal) cells, we used Student's t test. We analysed cell damage induced by SARS-CoV-2 and SARS-CoV with one-way ANOVA. Findings SARS-CoV-2 infected and replicated to comparable levels in human Caco2 cells and Calu3 cells over a period of 120 h (p=0·52). By contrast, SARS-CoV infected and replicated more efficiently in Caco2 cells than in Calu3 cells under the same multiplicity of infection (p=0·0098). SARS-CoV-2, but not SARS-CoV, replicated modestly in U251 (neuronal) cells (p=0·036). For animal species cell tropism, both SARS-CoV and SARS-CoV-2 replicated in non-human primate, cat, rabbit, and pig cells. SARS-CoV, but not SARS-CoV-2, infected and replicated in Rhinolophus sinicus bat kidney cells. SARS-CoV-2 consistently induced significantly delayed and milder levels of cell damage than did SARS-CoV in non-human primate cells (VeroE6, p=0·016; FRhK4, p=0·0004). Interpretation As far as we know, our study presents the first quantitative data for tropism, replication kinetics, and cell damage of SARS-CoV-2. These data provide novel insights into the lower incidence of diarrhoea, decreased disease severity, and reduced mortality in patients with COVID-19, with respect to the pathogenesis and high transmissibility of SARS-CoV-2 compared with SARS-CoV. Funding May Tam Mak Mei Yin, The Shaw Foundation Hong Kong, Richard Yu and Carol Yu, Michael Seak-Kan Tong, Respiratory Viral Research Foundation, Hui Ming, Hui Hoy and Chow Sin Lan Charity Fund, Chan Yin Chuen Memorial Charitable Foundation, Marina Man-Wai Lee, The Hong Kong Hainan Commercial Association South China Microbiology Research Fund, The Jessie & George Ho Charitable Foundation, Perfect Shape Medical, The Consultancy Service for Enhancing Laboratory Surveillance of Emerging Infectious Diseases and Research Capability on Antimicrobial Resistance for the Department of Health of the Hong Kong Special Administrative Region Government, The Theme-Based Research Scheme of the Research Grants Council, Sanming Project of Medicine in Shenzhen, and The High Level-Hospital Program, Health Commission of Guangdong Province, China.

Abstract: The widespread use of antibiotics in the past 80 years has saved millions of human lives, facilitated technological progress and killed incalculable numbers of microbes, both pathogenic and commensal. Human-associated microbes perform an array of important functions, and we are now just beginning to understand the ways in which antibiotics have reshaped their ecology and the functional consequences of these changes. Mounting evidence shows that antibiotics influence the function of the immune system, our ability to resist infection, and our capacity for processing food. Therefore, it is now more important than ever to revisit how we use antibiotics. This review summarizes current research on the short-term and long-term consequences of antibiotic use on the human microbiome, from early life to adulthood, and its effect on diseases such as malnutrition, obesity, diabetes, and Clostridium difficile infection. Motivated by the consequences of inappropriate antibiotic use, we explore recent progress in the development of antivirulence approaches for resisting infection while minimizing resistance to therapy. We close the article by discussing probiotics and fecal microbiota transplants, which promise to restore the microbiota after damage of the microbiome. Together, the results of studies in this field emphasize the importance of developing a mechanistic understanding of gut ecology to enable the development of new therapeutic strategies and to rationally limit the use of antibiotic compounds.

Abstract: We present a novel approach for unsupervised learning of depth and ego-motion from monocular video. Unsupervised learning removes the need for separate supervisory signals (depth or ego-motion ground truth, or multi-view video). Prior work in unsupervised depth learning uses pixel-wise or gradient-based losses, which only consider pixels in small local neighborhoods. Our main contribution is to explicitly consider the inferred 3D geometry of the whole scene, and enforce consistency of the estimated 3D point clouds and ego-motion across consecutive frames. This is a challenging task and is solved by a novel (approximate) backpropagation algorithm for aligning 3D structures. We combine this novel 3D-based loss with 2D losses based on photometric quality of frame reconstructions using estimated depth and ego-motion from adjacent frames. We also incorporate validity masks to avoid penalizing areas in which no useful information exists. We test our algorithm on the KITTI dataset and on a video dataset captured on an uncalibrated mobile phone camera. Our proposed approach consistently improves depth estimates on both datasets, and outperforms the state-of-the-art for both depth and ego-motion. Because we only require a simple video, learning depth and ego-motion on large and varied datasets becomes possible. We demonstrate this by training on the low quality uncalibrated video dataset and evaluating on KITTI, ranking among top performing prior methods which are trained on KITTI itself.1

Abstract: Atomically thin layers of two-dimensional materials can be assembled in vertical stacks that are held together by relatively weak van der Waals forces, enabling coupling between monolayer crystals with incommensurate lattices and arbitrary mutual rotation1,2. Consequently, an overarching periodicity emerges in the local atomic registry of the constituent crystal structures, which is known as a moire superlattice3. In graphene/hexagonal boron nitride structures4, the presence of a moire superlattice can lead to the observation of electronic minibands5–7, whereas in twisted graphene bilayers its effects are enhanced by interlayer resonant conditions, resulting in a superconductor–insulator transition at magic twist angles8. Here, using semiconducting heterostructures assembled from incommensurate molybdenum diselenide (MoSe2) and tungsten disulfide (WS2) monolayers, we demonstrate that excitonic bands can hybridize, resulting in a resonant enhancement of moire superlattice effects. MoSe2 and WS2 were chosen for the near-degeneracy of their conduction-band edges, in order to promote the hybridization of intra- and interlayer excitons. Hybridization manifests through a pronounced exciton energy shift as a periodic function of the interlayer rotation angle, which occurs as hybridized excitons are formed by holes that reside in MoSe2 binding to a twist-dependent superposition of electron states in the adjacent monolayers. For heterostructures in which the monolayer pairs are nearly aligned, resonant mixing of the electron states leads to pronounced effects of the geometrical moire pattern of the heterostructure on the dispersion and optical spectra of the hybridized excitons. Our findings underpin strategies for band-structure engineering in semiconductor devices based on van der Waals heterostructures9. Excitonic bands in MoSe2/WS2 heterostructures can hybridize, resulting in a resonant enhancement of moire superlattice effects.

Abstract: The Visual Object Tracking challenge 2015, VOT2015, aims at comparing short-term single-object visual trackers that do not apply pre-learned models of object appearance. Results of 62 trackers are presented. The number of tested trackers makes VOT 2015 the largest benchmark on short-term tracking to date. For each participating tracker, a short description is provided in the appendix. Features of the VOT2015 challenge that go beyond its VOT2014 predecessor are: (i) a new VOT2015 dataset twice as large as in VOT2014 with full annotation of targets by rotated bounding boxes and per-frame attribute, (ii) extensions of the VOT2014 evaluation methodology by introduction of a new performance measure. The dataset, the evaluation kit as well as the results are publicly available at the challenge website.

Abstract: Cells can detect and react to the biophysical properties of the extracellular environment through integrin-based adhesion sites and adapt to the extracellular milieu in a process called mechanotransduction. At these adhesion sites, integrins connect the extracellular matrix (ECM) with the F-actin cytoskeleton and transduce mechanical forces generated by the actin retrograde flow and myosin II to the ECM through mechanosensitive focal adhesion proteins that are collectively termed the “molecular clutch.” The transmission of forces across integrin-based adhesions establishes a mechanical reciprocity between the viscoelasticity of the ECM and the cellular tension. During mechanotransduction, force allosterically alters the functions of mechanosensitive proteins within adhesions to elicit biochemical signals that regulate both rapid responses in cellular mechanics and long-term changes in gene expression. Integrin-mediated mechanotransduction plays important roles in development and tissue homeostasis, and its dysregulation is often associated with diseases.

Abstract: We present the all-sky Planck catalogue of Sunyaev-Zeldovich (SZ) sources detected from the 29 month full-mission data. The catalogue (PSZ2) is the largest SZ-selected sample of galaxy clusters yet produced and the deepest all-sky catalogue of galaxy clusters. It contains 1653 detections, of which 1203 are confirmed clusters with identified counterparts in external data-sets, and is the first SZ-selected cluster survey containing > $10^3$ confirmed clusters. We present a detailed analysis of the survey selection function in terms of its completeness and statistical reliability, placing a lower limit of 83% on the purity. Using simulations, we find that the Y5R500 estimates are robust to pressure-profile variation and beam systematics, but accurate conversion to Y500 requires. the use of prior information on the cluster extent. We describe the multi-wavelength search for counterparts in ancillary data, which makes use of radio, microwave, infra-red, optical and X-ray data-sets, and which places emphasis on the robustness of the counterpart match. We discuss the physical properties of the new sample and identify a population of low-redshift X-ray under- luminous clusters revealed by SZ selection. These objects appear in optical and SZ surveys with consistent properties for their mass, but are almost absent from ROSAT X-ray selected samples.

Abstract: Policy gradient methods are an appealing approach in reinforcement learning because they directly optimize the cumulative reward and can straightforwardly be used with nonlinear function approximators such as neural networks. The two main challenges are the large number of samples typically required, and the difficulty of obtaining stable and steady improvement despite the nonstationarity of the incoming data. We address the first challenge by using value functions to substantially reduce the variance of policy gradient estimates at the cost of some bias, with an exponentially-weighted estimator of the advantage function that is analogous to TD(lambda). We address the second challenge by using trust region optimization procedure for both the policy and the value function, which are represented by neural networks. Our approach yields strong empirical results on highly challenging 3D locomotion tasks, learning running gaits for bipedal and quadrupedal simulated robots, and learning a policy for getting the biped to stand up from starting out lying on the ground. In contrast to a body of prior work that uses hand-crafted policy representations, our neural network policies map directly from raw kinematics to joint torques. Our algorithm is fully model-free, and the amount of simulated experience required for the learning tasks on 3D bipeds corresponds to 1-2 weeks of real time.

Abstract: We present an algorithm for fast global registration of partially overlapping 3D surfaces. The algorithm operates on candidate matches that cover the surfaces. A single objective is optimized to align the surfaces and disable false matches. The objective is defined densely over the surfaces and the optimization achieves tight alignment with no initialization. No correspondence updates or closest-point queries are performed in the inner loop. An extension of the algorithm can perform joint global registration of many partially overlapping surfaces. Extensive experiments demonstrate that the presented approach matches or exceeds the accuracy of state-of-the-art global registration pipelines, while being at least an order of magnitude faster. Remarkably, the presented approach is also faster than local refinement algorithms such as ICP. It provides the accuracy achieved by well-initialized local refinement algorithms, without requiring an initialization and at lower computational cost.

Abstract: AIMS: To investigate the work stress among Chinese nurses who are supporting Wuhan in fighting against Coronavirus Disease 2019 (COVID-19) infection and to explore the relevant influencing factors. BACKGROUND: The COVID-19 epidemic has posed a major threat to public health. Nurses have always played an important role in infection prevention, infection control, isolation, containment and public health. However, available data on the work stress among these nurses are limited. METHODS: A cross-sectional survey. An online questionnaire was completed by 180 anti-epidemic nurses from Guangxi. Data collection tools, including the Chinese version of the Stress Overload Scale (SOS) and the Self-rating Anxiety Scale (SAS), were used. Descriptive single factor correlation and multiple regression analyses were used in exploring the related influencing factors. RESULTS: The SOS (39.91 ± 12.92) and SAS (32.19 ± 7.56) scores of this nurse group were positively correlated (r = 0.676, p < .05). Multiple regression analysis showed that only children, working hours per week and anxiety were the main factors affecting nurse stress (p = .000, .048, .000, respectively). CONCLUSIONS: Nurses who fight against COVID-19 were generally under pressure. IMPLICATIONS FOR NURSING MANAGEMENT: Nurse leaders should pay attention to the work stress and the influencing factors of the nurses who are fighting against COVID-19 infection, and offer solutions to retain mental health among these nurses.

Abstract: Model pruning seeks to induce sparsity in a deep neural network's various connection matrices, thereby reducing the number of nonzero-valued parameters in the model. Recent reports (Han et al., 2015; Narang et al., 2017) prune deep networks at the cost of only a marginal loss in accuracy and achieve a sizable reduction in model size. This hints at the possibility that the baseline models in these experiments are perhaps severely over-parameterized at the outset and a viable alternative for model compression might be to simply reduce the number of hidden units while maintaining the model's dense connection structure, exposing a similar trade-off in model size and accuracy. We investigate these two distinct paths for model compression within the context of energy-efficient inference in resource-constrained environments and propose a new gradual pruning technique that is simple and straightforward to apply across a variety of models/datasets with minimal tuning and can be seamlessly incorporated within the training process. We compare the accuracy of large, but pruned models (large-sparse) and their smaller, but dense (small-dense) counterparts with identical memory footprint. Across a broad range of neural network architectures (deep CNNs, stacked LSTM, and seq2seq LSTM models), we find large-sparse models to consistently outperform small-dense models and achieve up to 10x reduction in number of non-zero parameters with minimal loss in accuracy.

Abstract: Since the publication of the Duchenne muscular dystrophy (DMD) care considerations in 2010, multidisciplinary care of this severe, progressive neuromuscular disease has evolved. In conjunction with improved patient survival, a shift to more anticipatory diagnostic and therapeutic strategies has occurred, with a renewed focus on patient quality of life. In 2014, a steering committee of experts from a wide range of disciplines was established to update the 2010 DMD care considerations, with the goal of improving patient care. The new care considerations aim to address the needs of patients with prolonged survival, to provide guidance on advances in assessments and interventions, and to consider the implications of emerging genetic and molecular therapies for DMD. The committee identified 11 topics to be included in the update, eight of which were addressed in the original care considerations. The three new topics are primary care and emergency management, endocrine management, and transitions of care across the lifespan. In part 1 of this three-part update, we present care considerations for diagnosis of DMD and neuromuscular, rehabilitation, endocrine (growth, puberty, and adrenal insufficiency), and gastrointestinal (including nutrition and dysphagia) management.

Abstract: Approximately 1.5 billion people worldwide are overweight or affected by obesity, and are at risk of developing type 2 diabetes, cardiovascular disease and related metabolic and inflammatory disturbances1,2. Although the mechanisms linking adiposity to associated clinical conditions are poorly understood, recent studies suggest that adiposity may influence DNA methylation3,4,5,6, a key regulator of gene expression and molecular phenotype7. Here we use epigenome-wide association to show that body mass index (BMI; a key measure of adiposity) is associated with widespread changes in DNA methylation (187 genetic loci with P < 1 × 10−7, range P = 9.2 × 10−8 to 6.0 × 10−46; n = 10,261 samples). Genetic association analyses demonstrate that the alterations in DNA methylation are predominantly the consequence of adiposity, rather than the cause. We find that methylation loci are enriched for functional genomic features in multiple tissues (P < 0.05), and show that sentinel methylation markers identify gene expression signatures at 38 loci (P < 9.0 × 10−6, range P = 5.5 × 10−6 to 6.1 × 10−35, n = 1,785 samples). The methylation loci identify genes involved in lipid and lipoprotein metabolism, substrate transport and inflammatory pathways. Finally, we show that the disturbances in DNA methylation predict future development of type 2 diabetes (relative risk per 1 standard deviation increase in methylation risk score: 2.3 (2.07–2.56); P = 1.1 × 10−54). Our results provide new insights into the biologic pathways influenced by adiposity, and may enable development of new strategies for prediction and prevention of type 2 diabetes and other adverse clinical consequences of obesity.

Abstract: How large are the benefits of transportation infrastructure projects, and what explains these benefits? To shed new light on these questions, I collect archival data from colonial India and use it to estimate the impact of India's vast railroad network. Guided by six predictions from a general equilibrium trade model, I find that railroads: (1) decreased trade costs and interregional price gaps; (2) increased interregional and international trade; (3) eliminated the responsiveness of local prices to local productivity shocks (but increased the transmission of these shocks between regions); (4) increased the level of real income (but harmed neighboring regions without railroad access); (5) decreased the volatility of real income; and (6), a sufficient statistic for the effect of railroads on welfare in the model accounts for virtually all of the observed reduced-form impact of railroads on real income. I find similar results from an instrumental variable specification, no spurious effects from over 40,000 km of lines that were approved but never built, and tight bounds on the estimated impact of railroads. These results suggest that transportation infrastructure projects can improve welfare significantly, and do so because they allow regions to exploit gains from trade.

Abstract: Knowledge distillation is a widely applicable technique for training a student neural network under the guidance of a trained teacher network. For example, in neural network compression, a high-capacity teacher is distilled to train a compact student; in privileged learning, a teacher trained with privileged data is distilled to train a student without access to that data. The distillation loss determines how a teacher's knowledge is captured and transferred to the student. In this paper, we propose a new form of knowledge distillation loss that is inspired by the observation that semantically similar inputs tend to elicit similar activation patterns in a trained network. Similarity-preserving knowledge distillation guides the training of a student network such that input pairs that produce similar (dissimilar) activations in the teacher network produce similar (dissimilar) activations in the student network. In contrast to previous distillation methods, the student is not required to mimic the representation space of the teacher, but rather to preserve the pairwise similarities in its own representation space. Experiments on three public datasets demonstrate the potential of our approach.

Abstract: We present EDA: easy data augmentation techniques for boosting performance on text classification tasks. EDA consists of four simple but powerful operations: synonym replacement, random insertion, random swap, and random deletion. On five text classification tasks, we show that EDA improves performance for both convolutional and recurrent neural networks. EDA demonstrates particularly strong results for smaller datasets; on average, across five datasets, training with EDA while using only 50% of the available training set achieved the same accuracy as normal training with all available data. We also performed extensive ablation studies and suggest parameters for practical use.

Abstract: Despite its importance to our understanding of physics at supranuclear densities, the equation of state (EoS) of matter deep within neutron stars remains poorly understood. Millisecond pulsars (MSPs) are among the most useful astrophysical objects in the Universe for testing fundamental physics, and place some of the most stringent constraints on this high-density EoS. Pulsar timing - the process of accounting for every rotation of a pulsar over long time periods - can precisely measure a wide variety of physical phenomena, including those that allow the measurement of the masses of the components of a pulsar binary system (Lorimer & Kramer 2005). One of these, called relativistic Shapiro delay (Shapiro 1964), can yield precise masses for both an MSP and its companion; however, it is only easily observed in a small subset of high-precision, highly inclined (nearly edge-on) binary pulsar systems. By combining data from the North American Nanohertz Observatory for Gravitational Waves (NANOGrav) 12.5-year data set with recent orbital-phase-specific observations using the Green Bank Telescope, we have measured the mass of the MSP J0740+6620 to be $\mathbf{2.14^{+0.10}_{-0.09}}$ solar masses (68.3% credibility interval; 95.4% credibility interval is $\mathbf{2.14^{+0.20}_{-0.18}}$ solar masses). It is highly likely to be the most massive neutron star yet observed, and serves as a strong constraint on the neutron star interior EoS.

Abstract: We propose a novel multi-task learning architecture, which allows learning of task-specific feature-level attention. Our design, the Multi-Task Attention Network (MTAN), consists of a single shared network containing a global feature pool, together with a soft-attention module for each task. These modules allow for learning of task-specific features from the global features, whilst simultaneously allowing for features to be shared across different tasks. The architecture can be trained end-to-end and can be built upon any feed-forward neural network, is simple to implement, and is parameter efficient. We evaluate our approach on a variety of datasets, across both image-to-image predictions and image classification tasks. We show that our architecture is state-of-the-art in multi-task learning compared to existing methods, and is also less sensitive to various weighting schemes in the multi-task loss function. Code is available at https://github.com/lorenmt/mtan.

Abstract: The invention features methods of diagnosis by assessing B7-H1 expression in a tissue from a subject that has, or is suspected of having, cancer, methods of treatment with agents that interfere with B7-H1-receptor interaction, methods of selecting candidate subjects likely to benefit from cancer immunotherapy, and methods of inhibiting expression of B7-H1.

Abstract: The Lomb-Scargle periodogram is a well-known algorithm for detecting and characterizing periodic signals in unevenly-sampled data. This paper presents a conceptual introduction to the Lomb-Scargle periodogram and important practical considerations for its use. Rather than a rigorous mathematical treatment, the goal of this paper is to build intuition about what assumptions are implicit in the use of the Lomb-Scargle periodogram and related estimators of periodicity, so as to motivate important practical considerations required in its proper application and interpretation.

Abstract: Purpose: Using gene-disrupted allogeneic T cells as universal effector cells provides an alternative and potentially improves current chimeric antigen receptor (CAR) T-cell therapy against cancers and infectious diseases.Experimental Design: The CRISPR/Cas9 system has recently emerged as a simple and efficient way for multiplex genome engineering. By combining lentiviral delivery of CAR and electro-transfer of Cas9 mRNA and gRNAs targeting endogenous TCR, β-2 microglobulin (B2M) and PD1 simultaneously, to generate gene-disrupted allogeneic CAR T cells deficient of TCR, HLA class I molecule and PD1.Results: The CRISPR gene-edited CAR T cells showed potent antitumor activities, both in vitro and in animal models and were as potent as non-gene-edited CAR T cells. In addition, the TCR and HLA class I double deficient T cells had reduced alloreactivity and did not cause graft-versus-host disease. Finally, simultaneous triple genome editing by adding the disruption of PD1 led to enhanced in vivo antitumor activity of the gene-disrupted CAR T cells.Conclusions: Gene-disrupted allogeneic CAR and TCR T cells could provide an alternative as a universal donor to autologous T cells, which carry difficulties and high production costs. Gene-disrupted CAR and TCR T cells with disabled checkpoint molecules may be potent effector cells against cancers and infectious diseases. Clin Cancer Res; 23(9); 2255-66. ©2016 AACR.

Abstract: Current state-of-the-art convolutional architectures for object detection are manually designed. Here we aim to learn a better architecture of feature pyramid network for object detection. We adopt Neural Architecture Search and discover a new feature pyramid architecture in a novel scalable search space covering all cross-scale connections. The discovered architecture, named NAS-FPN, consists of a combination of top-down and bottom-up connections to fuse features across scales. NAS-FPN, combined with various backbone models in the RetinaNet framework, achieves better accuracy and latency tradeoff compared to state-of-the-art object detection models. NAS-FPN improves mobile detection accuracy by 2 AP compared to state-of-the-art SSDLite with MobileNetV2 model in [32] and achieves 48.3 AP which surpasses Mask R-CNN [10] detection accuracy with less computation time.