Memorizing All for Implicit Discourse Relation Recognition

Implicit discourse relation recognition is a challenging task due to the absence of the necessary informative clues from explicit connectives. An implicit discourse relation recognizer has to carefully tackle the semantic similarity of sentence pairs and the severe data sparsity issue. In this article, we learn token embeddings to encode the structure of a sentence from a dependency point of view in their representations and use them to initialize a baseline model to make it really strong. Then, we propose a novel memory component to tackle the data sparsity issue by allowing the model to master the entire training set, which helps in achieving further performance improvement. The memory mechanism adequately memorizes information by pairing representations and discourse relations of all training instances, thus filling the slot of the data-hungry issue in the current implicit discourse relation recognizer. The proposed memory component, if attached with any suitable baseline, can help in performance enhancement. The experiments show that our full model with memorizing the entire training data provides excellent results on PDTB and CDTB datasets, outperforming the baselines by a fair margin.

Evaluation of Dynamic Load Reduction for a Tractor Semi-Trailer Using the Air Suspension System at all Axles of the Semi-Trailer

The air suspension system has become more and more popular in heavy vehicles and buses to improve ride comfort and road holding. This paper focuses on the evaluation of the dynamic load reduction at all axles of a semi-trailer with an air suspension system, in comparison with the one using a leaf spring suspension system on variable speed and road types. First, a full vertical dynamic model is proposed for a tractor semi-trailer (full model) with two types of suspension systems (leaf spring and air spring) for three axles at the semi-trailer, while the tractor’s axles use leaf spring suspension systems. The air suspension systems are built based on the GENSYS model; meanwhile, the remaining structural parameters are considered equally. The full model has been validated by experimental results, and closely follows the dynamical characteristics of the real tractor semi-trailer, with the percent error of the highest value being 6.23% and Pearson correlation coefficient being higher than 0.8, corresponding to different speeds. The survey results showed that the semi-trailer with the air suspension system can reduce the dynamic load of the entire field of speed from 20 to 100 km/h, given random road types from A to F according to the ISO 8608:2016 standard. The dynamic load coefficient (DLC) with the semi-trailer using the air spring suspension system can be reduced on average from 14.8% to 29.3%, in comparison with the semi-trailer using the leaf spring suspension system.

NETISCE: A Network-Based Tool for Cell Fate Reprogramming

The search for effective therapeutic targets in fields like regenerative medicine and cancer research has generated interest in cell fate reprogramming. This cellular reprogramming paradigm can drive cells to a desired target state from any initial state. However, methods for identifying reprogramming targets remain limited for biological systems that lack large sets of experimental data or a dynamical characterization. We present NETISCE, a novel computational tool for identifying cell fate reprogramming targets in static networks. NETISCE identifies reprogramming targets through the innovative use of control theory within a dynamical systems framework. Through validations in studies of cell fate reprogramming from developmental, stem cell, and cancer biology, we show that NETISCE can predict previously identified cell fate reprogramming targets and identify potentially novel combinations of targets. NETISCE extends cell fate reprogramming studies to larger-scale biological networks without the need for full model parameterization and can be implemented by experimental and computational biologists to identify parts of a biological system that are relevant for the desired reprogramming task.

A Comparison of Two-Stage Least Squares (TSLS) and Ordinary Least Squares (OLS) in Estimating the Structural Relationship between After-School Exercise and Academic Performance

The current study examines the structural relationship between the academic performance exam scores of Korean middle school students and their after-school exercise hours. Although prior literature theoretically or experimentally predicts that these variables are positively associated, this association is difficult to empirically verify without controlling for mutual effects with other variables, or unless a full model is estimated by specifying the whole structure of all variables affecting the two variables in question. Unlike previous studies, this study estimates the structural relationship using two-stage least squares method, which does not require experimental observations collected for our particular purpose or estimating the full model. From this estimation, we empirically affirm that there is a positive structural relationship between students’ after-school exercise hours and their academic performance exam scores, whereas the ordinary least squares method consistently estimates a negative relationship.

Rule-Based Models for Risk Estimation and Analysis of In-hospital Mortality in Emergency and Critical Care

We propose a novel method that uses associative classification and odds ratios to predict in-hospital mortality in emergency and critical care. Manual mortality risk scores have previously been used to assess the care needed for each patient and their need for palliative measures. Automated approaches allow providers to get a quick and objective estimation based on electronic health records. We use association rule mining to find relevant patterns in the dataset. The odds ratio is used instead of classical association rule mining metrics as a quality measure to analyze association instead of frequency. The resulting measures are used to estimate the in-hospital mortality risk. We compare two prediction models: one minimal model with socio-demographic factors that are available at the time of admission and can be provided by the patients themselves, namely gender, ethnicity, type of insurance, language, and marital status, and a full model that additionally includes clinical information like diagnoses, medication, and procedures. The method was tested and validated on MIMIC-IV, a publicly available clinical dataset. The minimal prediction model achieved an area under the receiver operating characteristic curve value of 0.69, while the full prediction model achieved a value of 0.98. The models serve different purposes. The minimal model can be used as a first risk assessment based on patient-reported information. The full model expands on this and provides an updated risk assessment each time a new variable occurs in the clinical case. In addition, the rules in the models allow us to analyze the dataset based on data-backed rules. We provide several examples of interesting rules, including rules that hint at errors in the underlying data, rules that correspond to existing epidemiological research, and rules that were previously unknown and can serve as starting points for future studies.

Validation of Pre-Transplantation Plasma ST2 Levels As a Prognostic Marker of 1-Year Non-Relapse Mortality after Allogeneic Hematopoietic Cell Transplantation

Introduction: Accurate assessment of the risk of non-relapse mortality (NRM) is important for making the shared decision about treatment with allogeneic hematopoietic cell transplantation (allo-HCT). We have shown that the pre-transplantation plasma level of suppression of tumorigenicity 2 (ST2)-a protein that is released to the bloodstream upon inflammation, cellular stress and endothelial damage-was associated with NRM after myeloablative allo-HCT [Gjærde et al., ASH Annual Meeting 2020, abstract #1524]. In an expanded cohort of both myeloablative- and non-myeloablative conditioned patients, we aimed to validate the value of pre-transplant ST2 in predicting 1-year NRM after allo-HCT. Methods: Pre-transplantation plasma ST2 levels were measured by enzyme-linked immunosorbent assays in 374 adult patients who underwent allo-HCT at Rigshospitalet between July 2015 and December 2019 (Table 1), using stored plasma samples collected at a median (Q1, Q3) of 23 (21, 24) days before allo-HCT. All patients were followed-up for at least 1 year after transplant. NRM was defined as all deaths in relapse-free patients. Given our sample size and outcome proportion, we could include four parameters in a logistic regression model of 1-year NRM to avoid severe overfitting [Riley et al., BMJ, 2020]. Based on our current clinical risk assessment practice, we included age (linear), comorbidity index (HCT-CI [Sorror et al., Blood, 2005], linear) and conditioning intensity (myeloablative vs. non-myeloablative) in a base model, to which we added the pre-transplantation ST2 level (linear) and assessed its incremental prognostic value [Steyerberg et al., Epidemiology, 2019]. The internal validity of the full model was estimated by bootstrapping [Steyerberg et al., J Clin Epidemiol, 2001]. Results: The median (Q1, Q3) pre-transplantation plasma ST2 level was 20.4 (15.2, 27.2) ng/mL. NRM at 1-year was 9% (N = 33). The main causes of NRM were organ failure (39%), infection (23%) and acute graft-versus-host disease (21%). Relapse risk at 1-year was 18%. The patients who constituted the 33 cases of 1-year NRM had a 2.7 ng/mL higher median pre-transplantation ST2 level than the remaining 341 patients (95% bootstrap confidence interval [CI] of the difference: -1.9, 6.2 ng/mL, Figure Panel A). In the full logistic regression model-including age, HCT-CI, conditioning intensity and ST2-ST2 was associated with 1-year NRM with an odds ratio of 1.32 (CI: 1.05, 1.65) per 10 ng/mL increase. Adding ST2 to the base model increased the model likelihood ratio χ 2 from 12.1 to 17.3 (p = 0.02), i.e. ST2 added a fraction of 30% (12.1/17.3) of new predictive information to age, HCT-CI and conditioning intensity. However, the ability of the full model to discriminate cases of NRM at 1-year remained poor with minimal improvement after adding ST2 (AUC up to 0.675 from 0.674 in the base model). The bootstrap-corrected AUC (the expected AUC of the full model used in a new population) was 0.63. Moreover, bootstrap-corrected estimates of predicted vs. observed risk revealed slight model miscalibration: lower predicted risks were generally underestimated, while higher predicted risks were overestimated (Figure Panel B). Conclusion: Pre-transplantation plasma levels of ST2 was a prognostic biomarker of 1-year NRM after allo-HCT, adding new predictive information to age, HCT-CI and conditioning intensity. However, internal validation of the full ST2-based prediction model revealed poor overall performance, precluding further validation and use of the model in clinical practice. When identifying prognostic biomarkers, investigation of overall predictive performance (in addition to already known prognostic factors) is needed before clinical usefulness can be evaluated. Figure 1 Figure 1. Disclosures No relevant conflicts of interest to declare.

Discrete epidemic models with two time scales

The main aim of the work is to present a general class of two time scales discrete-time epidemic models. In the proposed framework the disease dynamics is considered to act on a slower time scale than a second different process that could represent movements between spatial locations, changes of individual activities or behaviors, or others.To include a sufficiently general disease model, we first build up from first principles a discrete-time susceptible–exposed–infectious–recovered–susceptible (SEIRS) model and characterize the eradication or endemicity of the disease with the help of its basic reproduction number $\mathcal{R}_{0}$ R 0 .Then, we propose a general full model that includes sequentially the two processes at different time scales and proceed to its analysis through a reduced model. The basic reproduction number $\overline{\mathcal{R}}_{0}$ R ‾ 0 of the reduced system gives a good approximation of $\mathcal{R}_{0}$ R 0 of the full model since it serves at analyzing its asymptotic behavior.As an illustration of the proposed general framework, it is shown that there exist conditions under which a locally endemic disease, considering isolated patches in a metapopulation, can be eradicated globally by establishing the appropriate movements between patches.

The Violence of Leadership in Black Lives Matter

Since the murder of Trayvon Martin in 2012, the United States has seen the coalescing of black protestors and activists along with their multiracial collaborators under the banner of Movement for Black Lives (M4BL). This struggle against racialized violence, police brutality, and white supremacy has been witnessed in myriad ways, with two of its most prominent “reactions” occurring in Ferguson, Missouri, and Baltimore, Maryland. Within this struggle, the organization Black Lives Matter (BLM) has chosen to follow a “leader-full” model that replaces traditional hierarchical forms of leadership for that of collaboration and decentralization. This chapter thus seeks to highlight the competing notions of centralized and decentralized leadership within black liberation movements to better understand this model. Using the works of Barbara Ransby, Patrisse Khan-Cullors, and Frantz Fanon, this work will explore forms of black leadership that articulate alternative modes of accountability, service, and well-being within the struggle for black livability.