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.

MEMe: A Mutually Enhanced Modeling Method for Efficient and Effective Human Pose Estimation

In this paper, a mutually enhanced modeling method (MEMe) is presented for human pose estimation, which focuses on enhancing lightweight model performance, but with low complexity. To obtain higher accuracy, a traditional model scale is largely expanded with heavy deployment difficulties. However, for a more lightweight model, there is a large performance gap compared to the former; thus, an urgent need for a way to fill it. Therefore, we propose a MEMe to reconstruct a lightweight baseline model, EffBase transferred intuitively from EfficientDet, into the efficient and effective pose (EEffPose) net, which contains three mutually enhanced modules: the Enhanced EffNet (EEffNet) backbone, the total fusion neck (TFNeck), and the final attention head (FAHead). Extensive experiments on COCO and MPII benchmarks show that our MEMe-based models reach state-of-the-art performances, with limited parameters. Specifically, in the same conditions, our EEffPose-P0 with 256 × 192 can use only 8.98 M parameters to achieve 75.4 AP on the COCO val set, which outperforms HRNet-W48, but with only 14% of its parameters.

BJBN： BERT-JOIN-BiLSTM Networks for Medical Auxiliary Diagnostic

This study proposed a medicine auxiliary diagnosis model based on neural network. The model combines a bidirectional long short-term memory（Bi-LSTM）network and bidirectional encoder representations from transformers (BERT), which can well complete the extraction of local features of Chinese medicine texts. BERT can learn the global information of the text, so use BERT to get the global representation of medical text and then use Bi-LSTM to extract local features. We conducted a large number of comparative experiments on datasets. The results show that the proposed model has significant advantages over the state-of-the-art baseline model. The accuracy of the proposed model is 0.75.

LETR: An End-to-End Detector of Reconstruction Area in Blade’s Adaptive Machining with Transformer

In the leading/trailing edge’s adaptive machining of the near-net-shaped blade, a small portion of the theoretical part is retained for securing aerodynamic performance by manual work. However, this procedure is time-consuming and depends on the human experience. In this paper, we defined retained theoretical leading/trailing edge as the reconstruction area. To accelerate the reconstruction process, an anchor-free neural network model based on Transformer was proposed, named LETR (Leading/trailing Edge Transformer). LETR extracts image features from an aspect of mixed frequency and channel domain. We also integrated LETR with the newest meta-Acon activation function. We tested our model on the self-made dataset LDEG2021 on a single GPU and got an mAP of 91.9\%, which surpassed our baseline model, Deformable DETR by 1.1\%. Furthermore, we modified LETR’s convolution layer and named the new model after GLETR (Ghost Leading/trailing Edge Transformer) as a lightweight model for real-time detection. It is proved that GLETR has fewer weight parameters and converges faster than LETR with an acceptable decrease in mAP (0.1\%) by test results.

Predictive Value of Triglyceride-Glucose Index for In-hospital Mortality in Patients With Severe Fever With Thrombocytopenia Syndrome: A Multi-Center Observational Study

Background: Triglyceride-glucose (TyG) index has been proposed as a reliable indicator for insulin resistance and proved to be closely associated with the severity and mortality risk of infectious diseases. It remains indistinct whether TyG index performs an important role in predicting in-hospital mortality in patients with severe fever with thrombocytopenia syndrome (SFTS).Methods: The current study retrospectively recruited patients who were admitted for SFTS from January to December 2019 at five medical centers. TyG index was calculated in accordance with the description of previous study: Ln [fasting triglyceride (TG) (mg/dl) × fasting blood glucose (FBG) (mg/dl)/2]. The observational endpoint of the present study was defined as the in-hospital death.Results: In total, 79 patients (64.9 ± 10.5 years, 39.2% female) who met the enrollment criteria were enrolled in the current study. During the hospitalization period, 17 (21.5%) patients died in the hospital. TyG index remained a significant and independent predictor for in-hospital death despite being fully adjusted for confounders, either being taken as a nominal [hazard ratio (HR) 5.923, 95% CI 1.208–29.036, P = 0.028] or continuous (HR 7.309, 95% CI 1.854–28.818, P = 0.004) variate. TyG index exhibited a moderate-to-high strength in predicting in-hospital death, with an area under the receiver operating characteristic curve (AUC) of 0.821 (95% CI 0.712–0.929, P < 0.001). The addition of TyG index displayed significant enhancement on the predictive value for in-hospital death beyond a baseline model, manifested as increased AUC (baseline model: 0.788, 95% CI 0.676–0.901 vs. + TyG index 0.866, 95% CI 0.783–0.950, P for comparison = 0.041), increased Harrell's C-index (baseline model: 0.762, 95% CI 0.645–0.880 vs. + TyG index 0.813, 95% CI 0.724–0.903, P for comparison = 0.035), significant continuous net reclassification improvement (NRI) (0.310, 95% CI 0.092–0.714, P = 0.013), and significant integrated discrimination improvement (0.111, 95% CI 0.008–0.254, P = 0.040).Conclusion: Triglyceride-glucose index, a novel indicator simply calculated from fasting TG and FBG, is strongly and independently associated with the risk of in-hospital death in patients with SFTS.

SPC Mesoscale Analysis Compared to Field-Project Soundings: Implications for Supercell Environment Studies

Two-hundred-fifty-seven supercell proximity soundings obtained for field programs over the central U.S. are compared to profiles extracted from the SPC mesoscale analysis system (the SFCOA) to understand how errors in the SFCOA and in its baseline model analysis system – the RUC/RAP – might impact climatological assessments of supercell environments. A primary result is that the SFCOA underestimates the low-level storm-relative winds and wind shear, a clear consequence of the lack of vertical resolution near the ground. The near-ground (≤ 500 m) wind shear is underestimated similarly in near-field, far-field, tornadic, and nontornadic supercell environments. The near-ground storm-relative winds, however, are underestimated the most in the near field and in tornadic supercell environments. Under-prediction of storm-relative winds is therefore a likely contributor to the lack of differences in storm-relative winds between nontornadic and tornadic supercell environments in past studies that use RUC/RAP-based analyses. Furthermore, these storm-relative wind errors could lead to an under emphasis of deep-layer SRH variables relative to shallower SRH in discriminating nontornadic from tornadic supercells. The mean critical angles are 5–15° larger and farther from 90° in the observed soundings than in the SFCOA, particularly in the near field, likely indicating that the ratio of streamwise to crosswise horizontal vorticity is often smaller than that suggested by the SFCOA profiles. Errors in thermodynamic variables are less prevalent, but show low-level CAPE to be too low closer to the storms, a dry bias above the boundary layer, and the absence of shallow near-ground stable layers that are much more prevalent in tornadic supercell environments.

Direct-Contact Air, Gravel, Ground Heat Exchanger in Air Treatment Systems for Cowshed Air Conditioning

This paper describes the analysis of the possibility of use of the direct-contact air, gravel, ground heat exchanger (acronym GAHE), patented at the Wroclaw University of Science and Technology, as a means of improving microclimate parameters in dairy cows’ barns. Different possibilities of introducing GAHE to the standard mechanical ventilation system of cowsheds have been proposed and investigated. Based on literature data, the required air parameters in the barns of dairy cows were determined and discussed. Computer simulations were carried out and the results obtained were compared to the baseline model. Year-round changes in microclimate parameters, especially air temperature, relative humidity, and THI index were investigated. The benefits of GAHE use were indicated. The possible increase in the minimum air volume of ventilation during the winter season and the decrease in the maximum values of this parameter in the summer were presented. Indications were made of the systems where the application of GAHE could be the most beneficial. A further research path has been proposed.

Multipath Cross Graph Convolution for Knowledge Representation Learning

In the past, most of the entity prediction methods based on embedding lacked the training of local core relationships, resulting in a deficiency in the end-to-end training. Aiming at this problem, we propose an end-to-end knowledge graph embedding representation method. It involves local graph convolution and global cross learning in this paper, which is called the TransC graph convolutional network (TransC-GCN). Firstly, multiple local semantic spaces are divided according to the largest neighbor. Secondly, a translation model is used to map the local entities and relationships into a cross vector, which serves as the input of GCN. Thirdly, through training and learning of local semantic relations, the best entities and strongest relations are found. The optimal entity relation combination ranking is obtained by evaluating the posterior loss function based on the mutual information entropy. Experiments show that this paper can obtain local entity feature information more accurately through the convolution operation of the lightweight convolutional neural network. Also, the maximum pooling operation helps to grasp the strong signal on the local feature, thereby avoiding the globally redundant feature. Compared with the mainstream triad prediction baseline model, the proposed algorithm can effectively reduce the computational complexity while achieving strong robustness. It also increases the inference accuracy of entities and relations by 8.1% and 4.4%, respectively. In short, this new method can not only effectively extract the local nodes and relationship features of the knowledge graph but also satisfy the requirements of multilayer penetration and relationship derivation of a knowledge graph.

Bayesian Fatigue Life Prediction of Corroded Steel Reinforcing Bars

This paper presents a general method for fatigue life prediction of corroded steel reinforcing bars. A fatigue testing on standard specimens with pitting corrosion is carried out to obtain corrosion fatigue data. The maximum corrosion degree (MCD), characterizing the most severe site of the corrosion pit, is identified to have a log-linear relationship with the fatigue life. A fatigue life model incorporating the MCD and the stress range for corroded steel reinforcing bars is proposed. The model parameters are identified using the testing data, and the model is considered as the baseline model. To utilize the proposed model for life prediction of corroded steel reinforcing bars with different geometries and working conditions, the Bayesian method is employed to update the baseline model. The effectiveness of the overall method is demonstrated using independent datasets of realistic steel reinforcing bars.

Detection of Windthrown Tree Stems on UAV-Orthomosaics Using U-Net Convolutional Networks

The increasing number of severe storm events is threatening European forests. Besides the primary damages directly caused by storms, there are secondary damages such as bark beetle outbreaks and tertiary damages due to negative effects on the market. These subsequent damages can be minimized if a detailed overview of the affected area and the amount of damaged wood can be obtained quickly and included in the planning of clearance measures. The present work utilizes UAV-orthophotos and an adaptation of the U-Net architecture for the semantic segmentation and localization of windthrown stems. The network was pre-trained with generic datasets, randomly combining stems and background samples in a copy–paste augmentation, and afterwards trained with a specific dataset of a particular windthrow. The models pre-trained with generic datasets containing 10, 50 and 100 augmentations per annotated windthrown stems achieved F1-scores of 73.9% (S1Mod10), 74.3% (S1Mod50) and 75.6% (S1Mod100), outperforming the baseline model (F1-score 72.6%), which was not pre-trained. These results emphasize the applicability of the method to correctly identify windthrown trees and suggest the collection of training samples from other tree species and windthrow areas to improve the ability to generalize. Further enhancements of the network architecture are considered to improve the classification performance and to minimize the calculative costs.