Search

To mitigate the data transmission burden, an new memory event-trigger scheme is developed for the load frequency control (LFC) of multi-area networked power systems with the integration of various new energy sources. Hinfinitystability criteria under less conservative conditions are established by utilizing an improved Lyapunov stability theory and a second-order Bessel-Legendre (B-L) inequality for the LFC power system using the presented memory event-trigger scheme. Simulation on a two-area LFC system and IEEE-39 bus station test system are performed to validate the proposed method.

Burn injuries can result from mechanisms such as thermal, chemical, and electrical insults. A prompt and accurate assessment of burns is essential for deciding definitive clinical treatments. Currently, the primary approach for burn assessments, via visual and tactile observations, is approximately 60%-80% accurate. The gold standard is biopsy and a close second would be non-invasive methods like Laser Doppler Imaging (LDI) assessments, which have up to 97% accuracy in predicting burn severity and the required healing time. In this paper, we introduce a machine learning pipeline for assessing burn severities and segmenting the regions of skin that are affected by burn. Segmenting 2D colour images of burns allows for the injured versus non-injured skin to be delineated, clearly marking the extent and boundaries of the localized burn/region-of-interest, even during remote monitoring of a burn patient. We trained a convolutional neural network (CNN) to classify four severities of burns. We built a saliency mapping method, Boundary Attention Mapping (BAM), that utilises this trained CNN for the purpose of accurately localizing and segmenting the burn regions from skin burn images. We demonstrated the effectiveness of our proposed pipeline through extensive experiments and evaluations using two datasets; 1) A larger skin burn image dataset consisting of 1684 skin burn images of four burn severities, 2) An LDI dataset that consists of a total of 184 skin burn images with their associated LDI scans. The CNN trained using the first dataset achieved an average F1-Score of 78% and micro/macro- average ROC of 85% in classifying the four burn severities. Moreover, a comparison between the BAM results and LDI results for measuring injury boundary showed that the segmentations generated by our method achieved 91.60% accuracy, 78.17% sensitivity, and 93.37% specificity.

Additional file 1: Figure S1. Flowchart of patient selection. Table S1. Specification and emulation of a target trial evaluating the effect of NOACs versus warfarin on the risk of incident heart failure using real-world data from Taiwan’s NHIRD. Table S2. Baseline characteristics of elderly patients with atrial fibrillation and diabetes receiving NOAC or warfarin in the original population, without weighting or matching. Table S3. Risk of heart failure in elderly patients with atrial fibrillation and diabetes receiving NOAC versus warfarin in the sensitivity analysis applying propensity score matching and that applying multivariable regression models without propensity score methods. Table S4. Baseline characteristics of elderly patients with atrial fibrillation and diabetes receiving NOAC or warfarin in the population after propensity score matching.

Graphic abstract Droplet and aerosol transmission of COVID-19 are the most important concerns in dental clinics, due to the generation of large amounts of infected aerosol and droplets mixed with patient’s saliva during the procedures. The current approach to prevent airborne disease transmission is an extraoral aerosol suction unit: a stand-alone vacuum module with a segmented arm and cup. Despite the need for disease control in dental offices, these units are rarely seen due to the loud noise produced by vacuum, bulky size, and high cost. This paper describes the aerodynamic design optimization of an affordable, 3D printable, Extraoral Vacuum Aerosol Cup (EVAC) that can be directly connected to existing standard 7/16″ central vacuum high-volume evacuator (HVE) valves used for intraoral saliva absorption in a dental office. These HVEs are typically unsuitable for extraoral suction due to their low vacuum force. However, they can be used for extraoral suction, if the cup attachment is aerodynamically optimized for maximum suction efficiency. Fifteen different designs of EVAC are proposed and their suction processes were simulated with computational fluid dynamics. Droplets of various sizes are released to mimic the droplets produced during dental operation. The suction performances of EVACs with different sizes and shapes were compared to find out the designs with optimal performance. Prototypes of the optimized EVAC are 3D printed and tested at a dental office. Development and manufacturing of such a device will largely reduce the COVID-19 infection risk, thus improving the safety protection for both patients and doctors at dental offices.

The evaluation of abstractive summarization models typically uses test data that is identically distributed as training data. In real-world practice, documents to be summarized may contain input noise caused by text extraction artifacts or data pipeline bugs. The robustness of model performance under distribution shift caused by such noise is relatively under-studied. We present a large empirical study quantifying the sometimes severe loss in performance (up to 12 ROUGE-1 points) from different types of input noise for a range of datasets and model sizes. We then propose a light-weight method for detecting and removing such noise in the input during model inference without requiring any extra training, auxiliary models, or even prior knowledge of the type of noise. Our proposed approach effectively mitigates the loss in performance, recovering a large fraction of the performance drop, sometimes as large as 11 ROUGE-1 points.

Conditional language models are predominantly trained with maximum likelihood estimation (MLE), giving probability mass to sparsely observed target sequences. While MLE trained models assign high probability to plausible sequences given the context, the model probabilities often do not accurately rank-order generated sequences by quality. This has been empirically observed in beam search decoding as output quality degrading with large beam sizes, and decoding strategies benefiting from heuristics such as length normalization and repetition-blocking. In this work, we introduce sequence likelihood calibration (SLiC) where the likelihood of model generated sequences are calibrated to better align with reference sequences in the model's latent space. With SLiC, decoding heuristics become unnecessary and decoding candidates' quality significantly improves regardless of the decoding method. Furthermore, SLiC shows no sign of diminishing returns with model scale, and presents alternative ways to improve quality with limited training and inference budgets. With SLiC, we exceed or match SOTA results on a wide range of generation tasks spanning abstractive summarization, question generation, abstractive question answering and data-to-text generation, even with modest-sized models.

Machine learning algorithms typically assume independent and identically distributed samples in training and at test time. Much work has shown that high-performing ML classifiers can degrade significantly and provide overly-confident, wrong classification predictions, particularly for out-of-distribution (OOD) inputs. Conditional language models (CLMs) are predominantly trained to classify the next token in an output sequence, and may suffer even worse degradation on OOD inputs as the prediction is done auto-regressively over many steps. Furthermore, the space of potential low-quality outputs is larger as arbitrary text can be generated and it is important to know when to trust the generated output. We present a highly accurate and lightweight OOD detection method for CLMs, and demonstrate its effectiveness on abstractive summarization and translation. We also show how our method can be used under the common and realistic setting of distribution shift for selective generation (analogous to selective prediction for classification) of high-quality outputs, while automatically abstaining from low-quality ones, enabling safer deployment of generative language models., Published in ICLR 2023

OBJECTIVE Heart failure (HF) often develops in patients with diabetes and is recognized for its role in increased cardiovascular morbidity and mortality in this population. Most existing models predict risk in patients with prevalent rather than incident diabetes and fail to account for sex differences in HF risk factors. We derived sex-specific models in Ontario, Canada to predict HF at diabetes onset and externally validated these models in the U.K. RESEARCH DESIGN AND METHODS Retrospective cohort study using international population-based data. Our derivation cohort comprised all Ontario residents aged ≥18 years who were diagnosed with diabetes between 2009 and 2018. Our validation cohort comprised U.K. patients aged ≥35 years who were diagnosed with diabetes between 2007 and 2017. Primary outcome was incident HF. Sex-stratified multivariable Fine and Gray subdistribution hazard models were constructed, with death as a competing event. RESULTS A total of 348,027 Ontarians (45% women) and 54,483 U.K. residents (45% women) were included. At 1, 5, and 9 years, respectively, in the external validation cohort, the C-statistics were 0.81 (95% CI 0.79-0.84), 0.79 (0.77-0.80), and 0.78 (0.76-0.79) for the female-specific model; and 0.78 (0.75-0.80), 0.77 (0.76-0.79), and 0.77 (0.75-0.79) for the male-specific model. The models were well-calibrated. Age, rurality, hypertension duration, hemoglobin, HbA1c, and cardiovascular diseases were common predictors in both sexes. Additionally, mood disorder and alcoholism (heavy drinker) were female-specific predictors, while income and liver disease were male-specific predictors. CONCLUSIONS Our findings highlight the importance of developing sex-specific models and represent an important step toward personalized lifestyle and pharmacologic prevention of future HF development.

Brass (e.g., H80 brass) is widely used in different industrial sectors. To ensure its quality and reliability, the stress damage state of brass has to be monitored and analyzed in practice. This article introduces a new micro-magnetic method for monitoring and analyzing the tensile stress damage state of brass. Specifically, the stress-magnetic flux density model of H80 brass in the micro-magnetic environment was developed based on the stress-magnetic coupling effect. The magnetic flux density value was then used as an indicator of the different stages of tensile damage, and a new model of the spontaneous magnetic signal and structural deformation of H80 brass in the tensile process was established. The tensile test results show that the strain of H80 brass in the elastic deformation stage is 0%-5%, there is a small amount of dislocation, and the magnetic flux density does not change significantly; the strain increases to 5%-48% in the work-hardening stage, the dislocation density increases, a dislocation wall is formed at the internal lattice, and the magnetic flux density decreases with increasing tensile stress; the strain continues to increase to 48%-52% in the necking fracture stage and the internal lattice. When the strain at the necking fracture stage continues to increase to 48%-52%, the dislocation density of the internal lattice gradually stabilizes, accompanied by the appearance of sub-slip bands, and the number and density of slip bands increase; the magnetic flux density drops abruptly when fracturing occurs. These results show that the developed micro-magnetic method is accurate and reliable for monitoring and analyzing the tensile stress damage state of brass.

Widely used evaluation metrics for text generation either do not work well with longer texts or fail to evaluate all aspects of text quality. In this paper, we introduce a new metric called SMART to mitigate such limitations. Specifically, We treat sentences as basic units of matching instead of tokens, and use a sentence matching function to soft-match candidate and reference sentences. Candidate sentences are also compared to sentences in the source documents to allow grounding (e.g., factuality) evaluation. Our results show that system-level correlations of our proposed metric with a model-based matching function outperforms all competing metrics on the SummEval summarization meta-evaluation dataset, while the same metric with a string-based matching function is competitive with current model-based metrics. The latter does not use any neural model, which is useful during model development phases where resources can be limited and fast evaluation is required. Finally, we also conducted extensive analyses showing that our proposed metrics work well with longer summaries and are less biased towards specific models., Comment: code coming soon

While large pretrained Transformer models have proven highly capable at tackling natural language tasks, handling long sequence inputs continues to be a significant challenge. One such task is long input summarization, where inputs are longer than the maximum input context of most pretrained models. Through an extensive set of experiments, we investigate what model architectural changes and pretraining paradigms can most efficiently adapt a pretrained Transformer for long input summarization. We find that a staggered, block-local Transformer with global encoder tokens strikes a good balance of performance and efficiency, and that an additional pretraining phase on long sequences meaningfully improves downstream summarization performance. Based on our findings, we introduce PEGASUS-X, an extension of the PEGASUS model with additional long input pretraining to handle inputs of up to 16K tokens. PEGASUS-X achieves strong performance on long input summarization tasks comparable with much larger models while adding few additional parameters and not requiring model parallelism to train.

Supplementary Tables

Generative models have been applied in the medical imaging domain for various image recognition and synthesis tasks. However, a more controllable and interpretable image synthesis model is still lacking yet necessary for important applications such as assisting in medical training. In this work, we leverage the efficient self-attention and contrastive learning modules and build upon state-of-the-art generative adversarial networks (GANs) to achieve an attribute-aware image synthesis model, termed AttributeGAN, which can generate high-quality histopathology images based on multi-attribute inputs. In comparison to existing single-attribute conditional generative models, our proposed model better reflects input attributes and enables smoother interpolation among attribute values. We conduct experiments on a histopathology dataset containing stained H&E images of urothelial carcinoma and demonstrate the effectiveness of our proposed model via comprehensive quantitative and qualitative comparisons with state-of-the-art models as well as different variants of our model. Code is available at https://github.com/karenyyy/MICCAI2021AttributeGAN., MICCAI 2021

Adult-onset leukoencephalopathy with axonal spheroids (ALAS) is a group of hereditary, progressive, neurodegenerative disorders involving primarily the central nervous system white matter (WM). ALAS is characterized by patchy, asymmetrical myelin loss and axonal destruction in the WM, predominantly involving the frontoparietal regions. However, the asymmetrical and heterogenous involvement of different brain regions remains poorly characterized. In this study, digital spatial profiling was performed to investigate the region-specific expressions of 60 proteins. Conventional immunohistochemistry methods was used validate intrepretation of probes. Using a high-plex and high-throughput method, we provide evidence of regional heterogeneity in ALAS, particularly involving key markers of microglia composition, glial cell subtype and autophagy.

This study focuses on reachability problems in differential games. An improved level set method for computing reachable tubes is proposed in this paper. The reachable tube is described as a sublevel set of a value function, which is the viscosity solution of a Hamilton-Jacobi equation with running cost. We generalize the concept of reachable tubes and propose a new class of reachable tubes, which are referred to as cost-limited one. In particular, a performance index can be specified for the system, and a cost-limited reachable tube is a set of initial states of the system's trajectories that can reach the target set before the performance index increases to a given admissible cost. Such a reachable tube can be obtained by specifying the corresponding running cost function for the Hamilton-Jacobi equation. Different non-zero sublevel sets of the viscosity solution of the Hamilton-Jacobi equation at a certain time point can be used to characterize the cost-limited reachable tubes with different admissible costs (or the reachable tubes with different time horizons), thus reducing the storage space consumption. Several examples are provided to illustrate the validity and accuracy of the proposed method., 9 pages, 13 figures

The private Lookout Observatory (LO) monitored the classic nova V1112 Persei on 37 nights spanning over 80 days, beginning shortly after its discovery by Seiji Ueda on 25 Nov 2020. Images were captured at a high cadence, with exposure lengths of initially less than 2 seconds and with some sessions lasting more than ten hours. The standard error of the photometry was typically better than 5 thousandths of a magnitude (5 mmag). This cadence and precision allowed for not only the observation of the expected dimming of the nova, but also variability having a period of 0.608 +- 0.005 days. This data compliments the publicly available photometry from the American Association of Variable Star Observers (AAVSO) and the resultant data is combined to perform this photometric analysis. This paper does not attempt an in-depth physical analysis of the nova from an astrophysical perspective., Comment: 10 pages, 9 figures Accepted by JAAVSO

An important goal in human-robot-interaction (HRI) is for machines to achieve a close to human level of face perception. One of the important differences between machine learning and human intelligence is the lack of compositionality. This paper introduces a new scheme to enable generative adversarial networks to learn the distribution of face images composed of smaller parts. This results in a more flexible machine face perception and easier generalization to outside training examples. We demonstrate that this model is able to produce realistic high-quality face images by generating and piecing together the parts. Additionally, we demonstrate that this model learns the relations between the facial parts and their distributions. Therefore, the specific facial parts are interchangeable between generated face images.

The prognostic power of circulating cardiac biomarkers, their utility and pattern of release in coronavirus disease 2019 (COVID-19) patients have not been clearly defined. In this multi-centered retrospective study, we enrolled 3,219 patients with diagnosed COVID-19 admitted to 9 hospitals from December 31, 2019 to March 4, 2020, to estimate the associations and prognostic power of circulating cardiac injury markers with the poor outcomes of COVID-19. In the mixed-effect Cox model, after adjusting for age, gender and comorbidities, the adjusted hazard ratios of 28-day mortality for high-sensitivity cardiac troponin I (hs-cTnI) was 7.12 (95%CI, 4.60-11.03; P

Most prior work in the sequence-to-sequence paradigm focused on datasets with input sequence lengths in the hundreds of tokens due to the computational constraints of common RNN and Transformer architectures. In this paper, we study long-form abstractive text summarization, a sequence-to-sequence setting with input sequence lengths up to 100,000 tokens and output sequence lengths up to 768 tokens. We propose SEAL, a Transformer-based model, featuring a new encoder-decoder attention that dynamically extracts/selects input snippets to sparsely attend to for each output segment. Using only the original documents and summaries, we derive proxy labels that provide weak supervision for extractive layers simultaneously with regular supervision from abstractive summaries. The SEAL model achieves state-of-the-art results on existing long-form summarization tasks, and outperforms strong baseline models on a new dataset/task we introduce, Search2Wiki, with much longer input text. Since content selection is explicit in the SEAL model, a desirable side effect is that the selection can be inspected for enhanced interpretability.

Rationale: Use of angiotensin-converting enzyme inhibitors (ACEIs) and angiotensin II receptor blockers (ARBs) is a major concern for clinicians treating coronavirus disease 2019 (COVID-19) in patients with hypertension. Objective: To determine the association between in-hospital use of ACEI/ARB and all-cause mortality in COVID-19 patients with hypertension. Methods and Results: This retrospective, multi-center study included 1128 adult patients with hypertension diagnosed with COVID-19, including 188 taking ACEI/ARB (ACEI/ARB group; median age 64 [IQR 55-68] years; 53.2% men) and 940 without using ACEI/ARB (non-ACEI/ARB group; median age 64 [IQR 57-69]; 53.5% men), who were admitted to nine hospitals in Hubei Province, China from December 31, 2019 to February 20, 2020. Unadjusted mortality rate was lower in the ACEI/ARB group versus the non-ACEI/ARB group (3.7% vs. 9.8%; P = 0.01). In mixed-effect Cox model treating site as a random effect, after adjusting for age, gender, comorbidities, and in-hospital medications, the detected risk for all-cause mortality was lower in the ACEI/ARB group versus the non-ACEI/ARB group (adjusted HR, 0.42; 95% CI, 0.19-0.92; P =0.03). In a propensity score-matched analysis followed by adjusting imbalanced variables in mixed-effect Cox model, the results consistently demonstrated lower risk of COVID-19 mortality in patients who received ACEI/ARB versus those who did not receive ACEI/ARB (adjusted HR, 0.37; 95% CI, 0.15-0.89; P = 0.03). Further subgroup propensity score-matched analysis indicated that, compared to use of other antihypertensive drugs, ACEI/ARB was also associated with decreased mortality (adjusted HR, 0.30; 95%CI, 0.12-0.70; P = 0.01) in COVID-19 patients with hypertension. Conclusions: Among hospitalized COVID-19 patients with hypertension, inpatient use of ACEI/ARB was associated with lower risk of all-cause mortality compared with ACEI/ARB non-users. While study interpretation needs to consider the potential for residual confounders, it is unlikely that in-hospital use of ACEI/ARB was associated with an increased mortality risk.

Short term Metabolic Effects of Two Novel Oral Androgens, Dimethandrolone Undecanoate and 11β-MNTDC Supplemental File

Supplemental File

We propose an end-to-end neural model for zero-shot abstractive text summarization of paragraphs, and introduce a benchmark task, ROCSumm, based on ROCStories, a subset for which we collected human summaries. In this task, five-sentence stories (paragraphs) are summarized with one sentence, using human summaries only for evaluation. We show results for extractive and human baselines to demonstrate a large abstractive gap in performance. Our model, SummAE, consists of a denoising auto-encoder that embeds sentences and paragraphs in a common space, from which either can be decoded. Summaries for paragraphs are generated by decoding a sentence from the paragraph representations. We find that traditional sequence-to-sequence auto-encoders fail to produce good summaries and describe how specific architectural choices and pre-training techniques can significantly improve performance, outperforming extractive baselines. The data, training, evaluation code, and best model weights are open-sourced., first two authors contributed equally

International audience; Recent developments in cancer therapeutics have improved outcomes but have also been associated with cardiovascular complications. Therapies harnessing the immune system have been associated with an immune-mediated myocardial injury described as myocarditis. Immune checkpoint inhibitors are one such therapy with an increasing number of case and cohort reports describing a clinical syndrome of immune checkpoint inhibitor–associated myocarditis. Although the full spectrum of immune checkpoint inhibitor–associated cardiovascular disease still needs to be fully defined, described cases of myocarditis range from syndromes with mild signs and symptoms to fatal events. These observations in the clinical setting stand in contrast to outcomes from randomized clinical trials in which myocarditis is a rare event that is investigator reported and lacking in a specific case definition. The complexities associated with diagnosis, as well as the heterogeneous clinical presentation of immune checkpoint inhibitor–associated myocarditis, have made ascertainment and identification of myocarditis with high specificity challenging in clinical trials and other data sets, limiting the ability to better understand the incidence, outcomes, and predictors of these rare events. Therefore, establishing a uniform definition of myocarditis for application in clinical trials of cancer immunotherapies will enable greater understanding of these events. We propose an operational definition of cancer therapy-associated myocarditis that may facilitate case ascertainment and report and therefore may enhance the understanding of the incidence, outcomes, and risk factors of this novel clinical syndrome.

Discriminative neural networks offer little or no performance guarantees when deployed on data not generated by the same process as the training distribution. On such out-of-distribution (OOD) inputs, the prediction may not only be erroneous, but confidently so, limiting the safe deployment of classifiers in real-world applications. One such challenging application is bacteria identification based on genomic sequences, which holds the promise of early detection of diseases, but requires a model that can output low confidence predictions on OOD genomic sequences from new bacteria that were not present in the training data. We introduce a genomics dataset for OOD detection that allows other researchers to benchmark progress on this important problem. We investigate deep generative model based approaches for OOD detection and observe that the likelihood score is heavily affected by population level background statistics. We propose a likelihood ratio method for deep generative models which effectively corrects for these confounding background statistics. We benchmark the OOD detection performance of the proposed method against existing approaches on the genomics dataset and show that our method achieves state-of-the-art performance. We demonstrate the generality of the proposed method by showing that it significantly improves OOD detection when applied to deep generative models of images., Accepted to NeurIPS 2019

Context. Investigations of testosterone (T)-mediated feedback on LH secretion are few and contradictory. Discrepancies may reflect the choice of androgen, as well as dose, route and/or pattern (continuous vspulsatile) of administration. Moreover, for controlled T feedback studies, endogenous T secretion must be shut down and replaced by iv T infusion. Objective. The goal of this study was to quantify feedback in relation to the infused T pattern (pulsatile vs nonpulsatile) with the hypotheses that: 1) feedback by T pulses is less inhibitory, 2) T feedback operates partly at the pituitary level, 3) T feedback is controlled in part by body composition. Design. This was a placebo-controlled, blinded, and prospectively randomized crossover study comprising 16 men (age range 23-54 yr and BMI between 22.3-34.2 kg/m2). Subjects received ketoconazole (1600 mg over 22 h to block endogenous T secretion: serum T below 85 ng/dL) followed by pulsatile or continuous iv T addback (total T dose 5 mg/22 h). Cortisol inhibition was treated with low-dose dexamethasone. Setting. The study was performed in a Clinical Translational Research Unit. Interventions. Subjects underwent 14 h of blood sampling at 10-min intervals, starting at 0600 h. T was given continuously starting at 0900 h or as 45-min squarewave pulses every 90 min. To test pituitary feedback, a bolus iv injection of 33 ng/kg GnRH was given at 1630 h. Outcome measures. Log-transformed LH and T concentration ratios before and after GnRH administration. In addition, a four-parameter logistic regression between serum T concentration and LH secretion rate (calculated by deconvolution analysis) was used to independently estimate feedback of T on LH. Results. Despite higher T (total, bioavailable and free T) concentrations during pulsatile feedback, LH concentrations and secretion rates, whether driven by endogenous or exogenous GnRH, were similar to those during continuous infusion, indicating diminished pulsatile T feedback. This interpretation was corroborated by the right-shift (i.e. towards higher T concentrations) of the T-LH inhibitory dose-response curve. During continuous and pulsatile T infusion, LH concentrations and LH:T ratios depended positively on BMI to a similar extent before and after GnRH administration. Under both feedback regimens, basal but not pulsatile LH secretion correlated negatively with CT-estimated abdominal visceral fat mass. Age was not a significant covariate over the limited range explored. Conclusion. In summary, pulsatile T is less effective as a negative feedback signal on LH secretion than continuous T administration, as demonstrated by two independent methods. This difference in feedback strength was evident for exogenous GnRH-stimulated LH release as well, and thus occurs probably in part at the pituitary level, while not excluding additional hypothalamic involvement.

Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts all text-based language problems into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled data sets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new ``Colossal Clean Crawled Corpus'', we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our data set, pre-trained models, and code., Comment: Final version as published in JMLR

Recent work pre-training Transformers with self-supervised objectives on large text corpora has shown great success when fine-tuned on downstream NLP tasks including text summarization. However, pre-training objectives tailored for abstractive text summarization have not been explored. Furthermore there is a lack of systematic evaluation across diverse domains. In this work, we propose pre-training large Transformer-based encoder-decoder models on massive text corpora with a new self-supervised objective. In PEGASUS, important sentences are removed/masked from an input document and are generated together as one output sequence from the remaining sentences, similar to an extractive summary. We evaluated our best PEGASUS model on 12 downstream summarization tasks spanning news, science, stories, instructions, emails, patents, and legislative bills. Experiments demonstrate it achieves state-of-the-art performance on all 12 downstream datasets measured by ROUGE scores. Our model also shows surprising performance on low-resource summarization, surpassing previous state-of-the-art results on 6 datasets with only 1000 examples. Finally we validated our results using human evaluation and show that our model summaries achieve human performance on multiple datasets., Comment: Added results from mixed+stochastic model, test-set overlapping analysis; Code link added; Accepted for ICML 2020. arXiv admin note: text overlap with arXiv:1605.06560, arXiv:1205.2395, arXiv:0902.4351, arXiv:1610.09932, arXiv:nucl-ex/0512029 by other authors

11B-MNTDC 28 Day Study Supplemental File

Massively multi-label prediction/classification problems arise in environments like health-care or biology where very precise predictions are useful. One challenge with massively multi-label problems is that there is often a long-tailed frequency distribution for the labels, which results in few positive examples for the rare labels. We propose a solution to this problem by modifying the output layer of a neural network to create a Bayesian network of sigmoids which takes advantage of ontology relationships between the labels to help share information between the rare and the more common labels. We apply this method to the two massively multi-label tasks of disease prediction (ICD-9 codes) and protein function prediction (Gene Ontology terms) and obtain significant improvements in per-label AUROC and average precision for less common labels.

The full abstract for this thesis is available in the body of the thesis, and will be available when the embargo expires.

Clinicians spend a significant amount of time inputting free-form textual notes into Electronic Health Records (EHR) systems. Much of this documentation work is seen as a burden, reducing time spent with patients and contributing to clinician burnout. With the aspiration of AI-assisted note-writing, we propose a new language modeling task predicting the content of notes conditioned on past data from a patient's medical record, including patient demographics, labs, medications, and past notes. We train generative models using the public, de-identified MIMIC-III dataset and compare generated notes with those in the dataset on multiple measures. We find that much of the content can be predicted, and that many common templates found in notes can be learned. We discuss how such models can be useful in supporting assistive note-writing features such as error-detection and auto-complete., preprint

The driving force behind the recent success of LSTMs has been their ability to learn complex and non-linear relationships. Consequently, our inability to describe these relationships has led to LSTMs being characterized as black boxes. To this end, we introduce contextual decomposition (CD), an interpretation algorithm for analysing individual predictions made by standard LSTMs, without any changes to the underlying model. By decomposing the output of a LSTM, CD captures the contributions of combinations of words or variables to the final prediction of an LSTM. On the task of sentiment analysis with the Yelp and SST data sets, we show that CD is able to reliably identify words and phrases of contrasting sentiment, and how they are combined to yield the LSTM's final prediction. Using the phrase-level labels in SST, we also demonstrate that CD is able to successfully extract positive and negative negations from an LSTM, something which has not previously been done., Oral presentation at ICLR 2018

We show that generating English Wikipedia articles can be approached as a multi- document summarization of source documents. We use extractive summarization to coarsely identify salient information and a neural abstractive model to generate the article. For the abstractive model, we introduce a decoder-only architecture that can scalably attend to very long sequences, much longer than typical encoder- decoder architectures used in sequence transduction. We show that this model can generate fluent, coherent multi-sentence paragraphs and even whole Wikipedia articles. When given reference documents, we show it can extract relevant factual information as reflected in perplexity, ROUGE scores and human evaluations., Comment: Published as a conference paper at ICLR 2018

We performed a 28-day daily dosing pharmacokinetic and pharmacodynamic study of a novel oral androgen dimethandrolone undecanoate, DMAU, in healthy men. DMAU showed marked suppression of serum gonadotropins and testosterone in men, while being well-tolerated. There were no safety concerns in this 28-day study. Further studies are ongoing to assess the ability of DMAU to suppress spermatogenesis.

ive summarization has been studied using neural sequence transduction methods with datasets of large, paired document-summary examples. However, such datasets are rare and the models trained from them do not generalize to other domains. Recently, some progress has been made in learning sequence-to-sequence mappings with only unpaired examples. In our work, we consider the setting where there are only documents (product or business reviews) with no summaries provided, and propose an end-to-end, neural model architecture to perform unsupervised abstractive summarization. Our proposed model consists of an auto-encoder where the mean of the representations of the input reviews decodes to a reasonable summary-review while not relying on any review-specific features. We consider variants of the proposed architecture and perform an ablation study to show the importance of specific components. We show through automated metrics and human evaluation that the generated summaries are highly abstractive, fluent, relevant, and representative of the average sentiment of the input reviews. Finally, we collect a reference evaluation dataset and show that our model outperforms a strong extractive baseline., Comment: Accepted to ICML 2019

We performed a 28-day daily dosing pharmacokinetic and pharmacodynamic study of a novel oral androgen dimethandrolone undecanoate, DMAU, in healthy men. DMAU showed marked suppression of serum gonadotropins and testosterone in men, while being well-tolerated. There were no safety concerns in this 28-day study. Further studies are ongoing to assess the ability of DMAU to suppress spermatogenesis.

Safety and Pharmacokinetics of Oral 11βMNTDC - supplemental data

Background: Sacubitril/Valsartan has been shown to improve mortality and reduce hospitalizations in patients with heart failure with reduced ejection fraction (HFrEF). The effect of Sacubitril/Valsartan on ejection fraction (EF) and reverse remodeling parameters have not been previously described. Methods: We performed a single-center, retrospective, cohort study of HFrEF patients (n=48) who were treated with Sacubitril/Valsartan for a median duration of 3 months (Interquartile range 2-6 months). Clinical and echocardiographic parameters were reviewed at three time points (pre-baseline which was median of 18 months before starting Sacubitril/Valsartan, baseline before treatment started, and post-Sacubitril/Valsartan). Paired sample t-test and one-way repeated measures ANOVA were used for normally distributed data, while Wilcoxon Signed Rank test for non-normally distributed data. Results: Sacubitril/Valsartan use was associated with an average 5% (±1.2) increase in EF, from a mean baseline of 25.33% to 30.14% (p

Objective The objective of this study was to assess the risk of gadoxetate disodium in liver imaging for the development of nephrogenic systemic fibrosis (NSF) in patients with moderate to severe renal impairment. Materials and Methods We performed a prospective, multicenter, nonrandomized, open-label phase 4 study in 35 centers from May 2009 to July 2013. The study population consisted of patients with moderate to severe renal impairment scheduled for liver imaging with gadoxetate disodium. All patients received a single intravenous bolus injection of 0.025-mmol/kg body weight of liver-specific gadoxetate disodium. The primary target variable was the number of patients who develop NSF within a 2-year follow-up period. Results A total of 357 patients were included, with 85 patients with severe and 193 patients with moderate renal impairment, which were the clinically most relevant groups. The mean time period from diagnosis of renal disease to liver magnetic resonance imaging (MRI) was 1.53 and 5.46 years in the moderate and severe renal impairment cohort, respectively. Overall, 101 patients (28%) underwent additional contrast-enhanced MRI with other gadolinium-based MRI contrast agents within 12 months before the start of the study or in the follow-up. No patient developed symptoms conclusive of NSF within the 2-year follow-up. Conclusions Gadoxetate disodium in patients with moderate to severe renal impairment did not raise any clinically significant safety concern. No NSF cases were observed.

Recurrent neural network models with an attention mechanism have proven to be extremely effective on a wide variety of sequence-to-sequence problems. However, the fact that soft attention mechanisms perform a pass over the entire input sequence when producing each element in the output sequence precludes their use in online settings and results in a quadratic time complexity. Based on the insight that the alignment between input and output sequence elements is monotonic in many problems of interest, we propose an end-to-end differentiable method for learning monotonic alignments which, at test time, enables computing attention online and in linear time. We validate our approach on sentence summarization, machine translation, and online speech recognition problems and achieve results competitive with existing sequence-to-sequence models., ICML camera-ready version; 10 pages + 9 page appendix

Description: representative search strategy. (DOCX 13Â kb)

In this work we study metamaterial-enhanced graphene photodetectors operating in the mid-IR to THz. The detector element consists of a graphene ribbon embedded within a dual-metal split ring resonator, which acts like a cavity to enhance the absorption of electromagnetic radiation by the graphene ribbon, while the asymmetric metal contacts enable photothermoelectric detection. Detectors designed for the mid-IR demonstrate peak responsivity (referenced to total power) of ∼120 mV/W at 1500 cm–1 and are employed in the spectroscopic evaluation of vibrational resonances, thus demonstrating a key step toward a platform for integrated surface-enhanced sensing., ACS Photonics, 3 (6), ISSN:2330-4022

Additional file 3: Figure S1. Western blot analysis of light chain secretion.

ObjectiveAcutely restricting sleep worsens insulin sensitivity in healthy individuals whose usual sleep is normal in duration and pattern. The effect of recovery or weekend 'catch-up' sleep on insulin sensitivity and metabolically active hormones in individuals with chronic sleep restriction who regularly 'catch-up' on sleep at weekends is as yet unstudied.Design19 men (mean±SEM age 28·6±2·0years, BMI 26·0±0·8kg/m(2) ) with at least 6months' history (5·1±0·9years) of lifestyle-driven, restricted sleep during the working week (373±6·6min/night) with regular weekend 'catch-up' sleep (weekend sleep extension 37·4±2·3%) completed an in-laboratory, randomized, crossover study comprising two of three conditions, stratified by age. Conditions were 3weekend nights of 10 hours, 6 hours or 10hours time-in-bed with slow wave sleep (SWS) suppression using targeted acoustic stimuli.MeasurementsInsulin sensitivity was measured in the morning following the 3rd intervention night by minimal modelling of 19 samples collected during a 2-h oral glucose tolerance test. Glucose, insulin, c-peptide, leptin, peptide YY (PYY), ghrelin, cortisol, testosterone and luteinizing hormone (LH) were measured from daily fasting blood samples; HOMA-IR, HOMA-β and QUICKI were calculated.ResultsInsulin sensitivity was higher following three nights of sleep extension compared to sustained sleep restriction. Fasting insulin, c-peptide, HOMA-IR, HOMA-β, leptin and PYY decreased with 'catch-up' sleep, QUICKI and testosterone increased, while morning cortisol and LH did not change. Targeted acoustic stimuli reduced SWS by 23%, but did not alter insulin sensitivity.ConclusionsThree nights of 'catch-up' sleep improved insulin sensitivity in men with chronic, repetitive sleep restriction. Methods to improve metabolic health by optimizing sleep are plausible.

Multimodal biometric technology provides potential solutions for continuous user-to-device authentication in high-security mobile ad hoc networks (MANETs). This paper studies distributed combined authentication and intrusion detection with data fusion in such MANETs. Multimodal biometrics are deployed to work with intrusion detection systems (IDSs) to alleviate the shortcomings of unimodal biometric systems. Since each device in the network has measurement and estimation limitations, more than one device needs to be chosen, and observations can be fused to increase observation accuracy using Dempster-Shafer theory for data fusion. The system decides whether user authentication (or IDS input) is required and which biosensors (or IDSs) should be chosen, depending on the security posture. The decisions are made in a fully distributed manner by each authentication device and IDS. Simulation results are presented to show the effectiveness of the proposed scheme.

AimsUse of inotropic agents in patients with heart failure (HF) has been limited by adverse effects on outcomes. However, administration of positive inotropes at lower doses and concomitant treatment with beta-blockers might increase benefit-risk ratio. We investigated the effects of low doses of the positive inotrope enoximone on symptoms, exercise capacity, and major clinical outcomes in patients with advanced HF who were also treated with beta-blockers and other guideline-recommended background therapy.Methods and resultsThe Studies of Oral Enoximone Therapy in Advanced HF (ESSENTIAL) programme consisted of two identical, randomized, double-blind, placebo-controlled trials that differed only by geographic location (North and South America: ESSENTIAL-I; Europe: ESSENTIAL-II). Patients with New York Heart Association class III-IV HF symptoms, left ventricular ejection fraction < or = 30%, and one hospitalization or two ambulatory visits for worsening HF in the previous year were eligible for participation in the trials. The trials had three co-primary endpoints: (i) the composite of time to all-cause mortality or cardiovascular hospitalization, analysed in the two ESSENTIAL trials combined; (ii) the 6 month change from baseline in the 6 min walk test distance (6MWTD); and (iii) the Patient Global Assessment (PGA) at 6 months, both analysed in each trial separately. ESSENTIAL-I and -II randomized 1854 subjects at 211 sites in 16 countries. In the combined trials, all-cause mortality and the composite, first co-primary endpoint did not differ between the two treatment groups [hazard ratio (HR) 0.97; 95% confidence interval (CI) 0.80-1.17; and HR 0.98; 95% CI 0.86-1.12, respectively, for enoximone vs. placebo]. The two other co-primary endpoints were analysed separately in the two ESSENTIAL trials, as prospectively designed in the protocol. The 6MWTD increased with enoximone, compared with placebo, in ESSENTIAL-I (P = 0.025, not reaching, however, the pre-specified criterion for statistical significance of P < 0.020), but not in ESSENTIAL-II. No difference in PGA was observed in either trial.ConclusionAlthough low-dose enoximone appears to be safe in patients with advanced HF, major clinical outcomes are not improved.

Despite significant advances, cardiovascular disease is the leading cause of world-wide mortality, highlighting an important yet unmet clinical need. Understanding the pathophysiological basis underlying cardiovascular tissue injury and repair in therefore of prime importance. Following cardiac tissue injury, the immune system plays an important and complex role throughout the acute inflammatory response and regenerative response. This review will summarize the role of the immune system in cardiovascular disease, and focus on the idea that the immune system evolved to promote tissue homeostasis following tissue injury and/or infection, and that the inherent cost of this evolutionary development is unwanted inflammatory mediated damage. While inflammation induced tissue damage is of little evolutionary consequence in organisms that have limited life spans, as will be discussed below, inflammation plays a major role in the development of cardiovascular disease worldwide in humans.