Search

In recent years, a variety of supervised manifold learning techniques have been proposed to outperform their unsupervised alternative versions in terms of classification accuracy and data structure capturing. Some dissimilarity measures have been used in these techniques to guide the dimensionality reduction process. Their good performance was empirically demonstrated; however, the relevant analysis is still missing. This paper contributes to a theoretical analysis on a) how dissimilarity measures affect maintaining manifold neighbourhood structure and b) how supervised manifold learning techniques could contribute to the reduction of classification error. This paper also provides a cross-comparison between supervised and unsupervised manifold learning approaches in terms of structure capturing using Kendall’s Tau coefficients and co-ranking matrices. Four different metrics (including three dissimilarity measures and Euclidean distance) have been considered along with manifold learning methods such as Isomap, ${t}$ -Stochastic Neighbour Embedding ( ${t}$ -SNE), and Laplacian Eigenmaps (LE), in two datasets: Breast Cancer and Swiss-Roll. This paper concludes that although the dissimilarity measures used in the manifold learning techniques can reduce classification error, they do not learn well or preserve the structure of the hidden manifold in the high dimensional space, but instead, they destroy the structure of the data. Based on the findings of this paper, it is advisable to use supervised manifold learning techniques as a pre-processing step in classification. In addition, it is not advisable to apply supervised manifold learning for visualization purposes since the two-dimensional representation using supervised manifold learning does not improve the preservation of data structure.

A survey on automated lip-reading approaches is presented in this paper with the main focus being on deep learning related methodologies which have proven to be more fruitful for both feature extraction and classification. This survey also provides comparisons of all the different components that make up automated lip-reading systems including the audio-visual databases, feature extraction, classification networks and classification schemas. The main contributions and unique insights of this survey are: 1) A comparison of Convolutional Neural Networks with other neural network architectures for feature extraction; 2) A critical review on the advantages of Attention-Transformers and Temporal Convolutional Networks to Recurrent Neural Networks for classification; 3) A comparison of different classification schemas used for lip-reading including ASCII characters, phonemes and visemes, and 4) A review of the most up-to-date lip-reading systems up until early 2021.

In this paper, a neural network-based lip reading system is proposed. The system is lexicon-free and uses purely visual cues. With only a limited number of visemes as classes to recognise, the system is designed to lip read sentences covering a wide range of vocabulary and to recognise words that may not be included in system training. The system has been testified on the challenging BBC Lip Reading Sentences 2 (LRS2) benchmark dataset. Compared with the state-of-the-art works in lip reading sentences, the system has achieved a significantly improved performance with 15% lower word error rate. In addition, experiments with videos of varying illumination have shown that the proposed model has a good robustness to varying levels of lighting. The main contributions of this paper are: 1) The classification of visemes in continuous speech using a specially designed transformer with a unique topology; 2) The use of visemes as a classification schema for lip reading sentences; and 3) The conversion of visemes to words using perplexity analysis. All the contributions serve to enhance the accuracy of lip reading sentences. The paper also provides an essential survey of the research area.

As an alternative approach, viseme-based lipreading systems have demonstrated promising performance results in decoding videos of people uttering entire sentences. However, the overall performance of such systems has been significantly affected by the efficiency of the conversion of visemes to words during the lipreading process. As shown in the literature, the issue has become a bottleneck of such systems where the system’s performance can decrease dramatically from a high classification accuracy of visemes (e.g., over 90%) to a comparatively very low classification accuracy of words (e.g., only just over 60%). The underlying cause of this phenomenon is that roughly half of the words in the English language are homophemes, i.e., a set of visemes can map to multiple words, e.g., “time” and “some”. In this paper, aiming to tackle this issue, a deep learning network model with an Attention based Gated Recurrent Unit is proposed for efficient viseme-to-word conversion and compared against three other approaches. The proposed approach features strong robustness, high efficiency, and short execution time. The approach has been verified with analysis and practical experiments of predicting sentences from benchmark LRS2 and LRS3 datasets. The main contributions of the paper are as follows: (1) A model is developed, which is effective in converting visemes to words, discriminating between homopheme words, and is robust to incorrectly classified visemes; (2) the model proposed uses a few parameters and, therefore, little overhead and time are required to train and execute; and (3) an improved performance in predicting spoken sentences from the LRS2 dataset with an attained word accuracy rate of 79.6%—an improvement of 15.0% compared with the state-of-the-art approaches.

Objective: Monocyte to high-density lipoprotein ratio is considered as a new inflammatory marker and has been used to predict the severity of coronary heart disease and the incidence of adverse cardiovascular events (ACEs). However, there is a lack of data relative to large artery atherosclerosis (LAA) ischemic stroke. We investigated whether the monocyte to high-density lipoprotein (HDL) ratio (MHR) is related to the 3-month functional prognosis of LAA ischemic stroke.Materials and Methods: A retrospective analysis was conducted on 316 LAA ischemic stroke patients. The 3-month functional outcome was divided into good and poor according to the modified Rankin Scale (mRS) score. Multivariate logistic regression analysis was performed to evaluate the correlation between MHR and prognosis of ischemic stroke.Results: The MHR level of poor functional outcome group was higher than that of the good functional outcome group [0.44 (0.3, 0.55) vs. 0.38 (0.27, 0.5), P = 0.025]. Logistic stepwise multiple regression revealed that MHR [odds ratio (OR) 9.464, 95%CI 2.257–39.678, P = 0.002] was an independent risk factor for the 3-month poor outcome of LAA ischemic stroke. Compared to the lower MHR tertile, the upper MHR tertile had a 3.03-fold increase (95% CI 1.475–6.225, P = 0.003) in the odds of poor functional outcome after adjustment for potential confounders. Moreover, a multivariable-adjusted restricted cubic spline (RCS) showed a positive close to a linear pattern of this association.Conclusion: Elevated MHR was independently associated with an increased risk of poor 3-month functional outcome of patients with LAA ischemic stroke.

Purely data-driven methods often fail to learn accurate conditional probability table (CPT) parameters of discrete Bayesian networks (BNs) when training data are scarce or incomplete. A practical and efficient means of overcoming this problem is to introduce qualitative parameter constraints derived from expert judgments. To exploit such knowledge, in this paper, we provide a constrained maximum a posteriori (CMAP) method to learn CPT parameters by incorporating convex constraints. To further improve the CMAP method, we present a type of constrained Bayesian Dirichlet priors that is compatible with the given constraints. Combined with the CMAP method, we propose an improved expectation maximum algorithm to process incomplete data. Experiments are conducted on learning standard BNs from complete and incomplete data. The results show that the proposed method outperforms existing methods, especially when data are extremely limited or incomplete. This finding suggests the potential effective application of CMAP to real-world problems. Moreover, a real facial action unit (AU) recognition case with incomplete data is conducted by applying different parameter learning methods. The results show that the recognition accuracy of respective recognition methods can be improved by the AU BN, which is trained by the proposed method.

Traumatic brain injury (TBI) is one of the most common causes of neurological damage in young human populations. Vitamin B12 has been reported to promote axon growth of neuronal cells after peripheral nerve injury, which is currently used for the treatment of peripheral nerve damage in the clinical trial. Thus, we hypothesized that TBI can be attenuated by vitaminB12 treatment through its beneficial role on axon regeneration after nerve injury. To confirm it, the biological function of vitaminB12 was characterized using hematoxylin and eosin (H&E) staining, Luxol fast blue (LFB) staining, western blot analysis, and immunohistochemistry staining. The results showed that the neurological functional recovery was improved in the VitaminB12-treated group after TBI, which may be due to downregulation of the endoplasmic reticulum stress-related apoptosis signaling pathway. Moreover, the microtubule stabilization, remyelination and myelin reparation were rescued by vitamin B12, which was consistent with the treatment of 4-phenylbutyric acid (4-PBA), an endoplasmic reticulum stress inhibitor. The study suggests that vitamin B12 may be useful as a novel neuroprotective drug for TBI.

We present our latest research work on the development of a skin image analysis tool by using machine-learning algorithms. Skin imaging is very import in skin research. Over the years, we have used and developed different types of skin imaging techniques. As the number of skin images and the type of skin images increase, there is a need of a dedicated skin image analysis tool. In this paper, we report the development of such software tool by using the latest MATLAB App Designer. It is simple, user friendly and yet powerful. We intend to make it available on GitHub, so that others can benefit from the software. This is an ongoing project, we are reporting here what we have achieved so far, and more functions will be added to the software in the future.

BackgroundThe COVID-19 virus is an emerging virus rapidly spread worldwide This study aimed to establish an effective diagnostic nomogram for suspected COVID-19 pneumonia patients.METHODSWe used the LASSO aggression and multivariable logistic regression methods to explore the predictive factors associated with COVID-19 pneumonia, and established the diagnostic nomogram for COVID-19 pneumonia using multivariable regression. This diagnostic nomogram was assessed by the internal and external validation data set. Further, we plotted decision curves and clinical impact curve to evaluate the clinical usefulness of this diagnostic nomogram.RESULTSThe predictive factors including the epidemiological history, wedge- shaped or fan-shaped lesion parallel to or near the pleura, bilateral lower lobes, ground glass opacities, crazy paving pattern and white blood cell (WBC) count were contained in the nomogram. In the primary cohort, the C-statistic for predicting the probability of the COVID-19 pneumonia was 0.967, even higher than the C-statistic (0.961) in initial viral nucleic acid nomogram which was established using the univariable regression. The C-statistic was 0.848 in external validation cohort. Good calibration curves were observed for the prediction probability in the internal validation and external validation cohort. The nomogram both performed well in terms of discrimination and calibration. Moreover, decision curve and clinical impact curve were also beneficial for COVID- 19 pneumonia patients.CONCLUSIONOur nomogram can be used to predict COVID-19 pneumonia accurately and favourably.

Background: This study aimed to establish an effective diagnostic nomogram for suspected SARS-CoV-2 pneumonia patients br br Methods: We used the LASSO aggre

In this paper, a deep learning-based approach is applied to high dimensional, high-volume, and high-sparsity medical data to identify critical casual attributions that might affect the survival of a breast cancer patient. The Surveillance Epidemiology and End Results (SEER) breast cancer data is explored in this study. The SEER data set contains accumulated patient-level and treatment-level information, such as cancer site, cancer stage, treatment received, and cause of death. Restricted Boltzmann machines (RBMs) are proposed for dimensionality reduction in the analysis. RBM is a popular paradigm of deep learning networks and can be used to extract features from a given data set and transform data in a non-linear manner into a lower dimensional space for further modelling. In this study, a group of RBMs has been trained to sequentially transform the original data into a very low dimensional space, and then the k-means clustering is conducted in this space. Furthermore, the results obtained about the cluster membership of the data samples are mapped back to the original sample space for interpretation and insight creation. The analysis has demonstrated that essential features relating to breast cancer survival can be effectively extracted and brought forward into a much lower dimensional space formed by RBMs.

This paper presents a systematic work aiming to improve the efficiency of task processing in a networked UAV combat cloud system. The work consists of three major aspects: (1) an architecture of UAV combat cloud systems—such a system provides the necessary resource pool for powerful computing and storage facilities and defines the attributes of the entities in the resource pool in detail; (2) an online adaptive task redistribution and scheduling algorithm—the algorithm involves task migration being performed on virtual machines on the cloud system and aims to address the problems caused by static task scheduling approaches; and (3) an online virtual machine and task migration algorithm—the algorithm considers collectively the priority type and quantity of the tasks to be migrated on virtual machines along with time constraints to determine the migration of virtual machine or task and optimize resource usages. Experimental simulation results have demonstrated that the proposed system and the relevant algorithms can significantly improve the efficiency of task schedule.

Skin hydration measurement is very important for many clinical studies. Skin capacitive imaging is a novel technique that can be used for in-vivo skin hydration measurements. It is based on permittivity measurement principle, and can generate a skin water content image using a capacitive matrix sensor. In this paper, we present our latest study on the skin capacitive imaging analysis using Deep Learning GoogLeNet. We also present the development of a graphical user interface programme in order to simply the usage and improve the efficiency. The skin capacitive images are divided into three groups according to volunteers (V1, V2, V3), gender (male and female), and skin sites (face, forearm, forehead, neck, palm, and lower leg). GoogLeNet is used for image classifications. The results show that GoogLeNet can effectively differentiate the different skin capacitive images from different categories. We will first present the skin capacitive imaging technology and then present the experimental results.

Traumatic brain injury (TBI) caused by an external mechanical force acting on the brain is a serious neurological condition. Inflammation plays an important role in prolonging secondary tissue injury after TBI, leading to neuronal cell death and dysfunction. Isoliquiritigenin (ILG) is a flavonoid monomer with anti-inflammatory characteristic. Thus, we had investigated the potential protective effects of ILG on TBI-induced injuries and identified the mechanisms underlying it. Here, we have demonstrated that ILG preserves blood brain barrier (BBB) integrity in vivo, suppresses the activation of microglia and inflammatory responses in mice after TBI, consequently leading to neurofunctional deficits, brain oedema, structural damage, and macrophage infiltration. In vitro, ILG exerts anti-inflammatory effect, and upregulates tight junction proteins 120‑β‑catenin and occludin in SH‑SY5Y cells under oxygen glucose deprivation/reoxygenation (OGD/D) condition. Additionally, we found that PI3K/AKT/GSK‑3β signalling pathway is involved in ILG treatment for TBI. To further confirm it, we had used SC79 (ethyl 2‑amino‑6‑chloro‑4‑(1‑cyano‑2‑ethoxy‑2‑oxoethyl)‑4H‑chromene‑3‑carboxylate), an Akt specific activator, to activate Akt, we found that SC79 partially reduces the protective effect of ILG for TBI. Overall, our current study reveals the neuroprotective role of ILG on TBI-induced BBB damage, downregulated tight junction proteins via PI3K/AKT/GSK‑3β signalling pathway. Furthermore, ILG suppresses the secretion of pro-inflammatory cytokines after TBI through inhibiting the PI3K/AKT/GSK‑3β/NF‑κB signalling pathway. Our findings suggest that GSK‑3β is a key regulatory factor during TBI-induced secretion of inflammatory cytokines, neuronal apoptosis and destruction of BBB.

Traumatic brain injury (TBI) is one of the most serious public health problems in the world. TBI causes neurological deficits by triggering secondary injuries. Hydrogen sulfide (H2S), a gaseous mediator, has been reported to exert neuroprotective effects in central nervous system diseases, such as TBI. However, the molecular mechanisms involved in this effect are still unclear. The present study was designed to explore the ability of NaHS, a H2S donor, to provide neuroprotection in a mouse model of TBI and to discover the associated molecular mechanisms of these protective effects. Here, we found that administration of NaHS not only maintained the integrity of the blood brain barrier (BBB), protected neurons from apoptosis, and promoted remyelination and axonal reparation but also protected mitochondrial function. In addition, we found that autophagy was inhibited after treatment with NaHS following TBI, an effect that was induced by activation of the PI3K/AKT/mTOR signalling pathway. Our study indicated that H2S treatment is beneficial for TBI, pointing to H2S as a potential therapeutic target for treating TBI.

This study aims to assess the diagnostic and prognostic value of propionic acid in patients with septic shock on a medical intensive care unit (ICU).Serum propionic acid and clinical common cytokines levels were measured within 24h after the diagnosis of sepsis, and the Acute Physiology and Chronic Health Evaluation II (APACHE II) score, Sequential Organ Failure Assessment (SOFA) score, and Mortality were recorded in ICU. A 28-day and 90-day follow-up was performed for all patients.A total of 118 septic patients were enrolled in this study. The propionic acid was higher in patients with septic shock compared with sepsis. Multivariate logistic regression analysis showed that propionic acid was independent predictor of sepsis (odds ratio: 1.279; 95% confidence interval: 1.069-1.530; P=0.007) and septic shock (odds ratio: 1.859; 95% confidence interval: 1.342-2.576; P0.001) and ICU mortality (odds ratio: 1.331; 95% confidence interval: 1.107-1.600; P=0.002), 28-day mortality (odds ratio: 1.259; 95% confidence interval: 1.046-1.514; P=0.015) and 90-day mortality (odds ratio: 1.304; 95% confidence interval: 1.092-1.558; P=0.003). The receiver operating characteristic curve (AUC) analysis showed the areas under of propionic acid on ICU admission day for predicting sepsis and septic shock were 0.773 and 0.85 respectively, the areas under of propionic acid for predicting ICU mortality, 28-d and 90-d mortality were 0.779, 0.739 and 0.809 respectively. Using a PA cutoff of 0.053 and 0.095 for predicting sepsis and septic shock respectively, the sensitivity was 97.62% and 85.5%, and the specificity was 58% and 83.5%, respectively. Using a PA cutoff of 0.139 for predicting ICU mortality, 28- and 90-day mortality, the sensitivity was 69.39%, 67.44% and 69.09% respectively, and the specificity was 78.26%, 73.33% and 82.54% respectively.Propionic acid showed diagnostic capacity to diagnose septic shock and revealed prognostic information for mortality.

To improve the learning accuracy of the parameters in a Bayesian network from a small data set, domain knowledge is normally incorporated into the learning process as parameter constraints. MAP-based (Maximum a Posteriori) methods that utilize both sample data and domain knowledge have been well studied in the literature. Among all the MAP-based methods, the QMAP (Qualitatively Maximum a Posteriori) method is one of the algorithms with the highest learning performance. When the data is insufficient, however, the estimation given by the QMAP often fails to satisfy all the parameter constraints, and this has made the overall QMAP estimation unreliable. To ensure that a QMAP estimation does not violate any given parameter constraints and further to improve the learning accuracy, a FC-QMAP (Further Constrained Qualitatively Maximum a Posteriori) algorithm is proposed in this paper. The algorithm regulates QMAP estimation by replacing data estimation with a further constrained estimation via convex optimization. Experiments and theoretical analysis show that the proposed algorithm outperforms most of the existing parameter learning methods including Maximum Likelihood, Constrained Maximum Likelihood, Maximum Entropy, Constrained Maximum Entropy, Maximum a Posteriori, and Qualitatively Maximum a Posteriori.

Aiming at the maneuvering decision of aircraft in air combat, an intelligent maneuvering decision model based on convolutional neural network(CNN) is proposed in this paper. Firstly, the situation data, maneuvering decision variables and evaluation indexs are given, and a CNN model that can realize intelligent maneuvering decision is established. Then, according to the evaluation indexes, the structure and parameters of the CNN model are adjusted through the simulation experiments to improve the accuracy and robustness of the maneuvering decision. After that, the validity of the intelligent maneuvering decision model proposed in this paper is verified through comparative experiments that the CNN model can make stable maneuvering decisions with high accuracy. Finally, the flight path in an air combat process is presented.

Blood analysis is an indispensable means of detection in criminal investigation, customs security and quarantine, anti-poaching of wildlife, and other incidents. Detecting the species of blood is one of the most important analyses. In order to classify species by analyzing Raman spectra of blood, a recognition method based on deep learning principle is proposed in this paper. This method can realize multi-identification blood species, by constructing a one-dimensional convolution neural network and establishing a Raman spectra database containing 20 kinds of blood. The network model is obtained through training, and then is employed to predict the testing set data. The average accuracy of blind detection is more than 97%. In this paper, we try to increase the diversity of data to improve the robustness of the model, optimize the network and adjust the hyperparameters to improve the recognition ability of the model. The evaluation results show that the deep learning model has high recognition performance to distinguish the species of blood.

The article presents a study to design a 6-degree of freedom robotic arm which can pick up objects in random positions on a 2D surface based on Arduino microcontroller, ultrasonic sensors and picamera. The robotic arm is able to recognise objects based on computer vision algorithm for shape detection. The ultrasonic sensor measures the distance between the objects and the robotic arm, and the position of the objects in the real world will be detected by its mass centre in the image to improve the accuracy of the pick-up movement. Arduino microcontroller will calculate the rotation angles for the joints of the robotic arm by using inverse kinematics algorithms. The movement of the robotic arm also can be controlled by an Amazon Alexa voice assistance device. The experiment of applying the artificial neural network to control the robotic arm pick-up movement is achieved. The artificial neural network can manipulate the position of the robotic arm to pick up objects after training using the values which are calculated by inverse kinematics equations. The Raspberry Pi is used for processing the computer vision data, and voice commands from Alexa Voice Service based on cloud service.

The intelligent flight control of the aircraft is the key process in the air combat maneuver process. The traditional flight control method has many steps, long time and low precision, which have great drawbacks in the air combat process. In this paper, based on the background of deep learning, a flight control model based on autoencoder is proposed. Using the characteristics of autoencoder dimension reduction and feature extraction, the low-dimensional attitude parameters of high-dimensional aircraft can be extracted from high-dimensional flight attitude parameters. The eigenvalues are then automatically obtained through the neural network to change the attitude control of the aircraft. In this paper, the basic framework and training methods of the model are designed, and the influence of various parameters of the autoencoder network on the performance of the model is deeply studied. The experimental results show that the proposed model has better prediction accuracy and convergence performance than the traditional BP neural network, and achieves the purpose of intelligently and quickly obtaining flight attitude control to intelligently control aircraft flight.

In this paper, a novel machine learning based approach is proposed for automated cost analysis from priced bill of quantities prepared by tenders in the construction industry. The proposed approach features: 1) An effective integration of structured project-specific information with surveyor’s domain knowledge in order to model the complex interrelationships between the specifications and descriptions of an item and its trade category; 2) An effective transformation to map the original data into a 2-dimensional space to tackle issues of high dimensionality in modelling, and 3) Simply classifiers with good classification capability. Relevant comparative experimental results have demonstrated the effectiveness of the proposed approach.

Traumatic brain injury (TBI) is among the leading causes of irreversible neurological damage and death worldwide. The aim of the present study was to investigate whether edaravone (EDA) had a neuroprotective effect on TBI as well as to identify the potential mechanism. Results demonstrated that EDA suppressed inflammatory and oxidative responses in mice following TBI. This was evidenced by a reduction in glutathione peroxidase, interleukin 6, tumor necrosis factor-α and hydrogen peroxide levels, in addition to an increase in hemeoxygenase-1, quinone oxidoreductase 1 and superoxide dismutase levels, thereby mitigating neurofunctional deficits, cell apoptosis and structural damage. EDA prevented the transfer of NF-κB protein from the cytoplasm to the nucleus, whilst promoting the expression of nuclear factor erythroid 2-related factor 2 (Nrf2) protein in mice following TBI. These results indicated that EDA exerted neuroprotective effects, including impeding neurofunctional deficits, cell apoptosis and structural damage, in mice with TBI, potentially via suppression of NF-κB-mediated inflammatory activation and promotion of the Nrf2 antioxidant pathway.

In recent years, many studies have focused on autophagy, an evolutionarily conserved mechanism that relies on lysosomes to achieve cellular metabolic requirements and organelle turnover, and revealed its important role in animal models of traumatic injury. Autophagy is a double-edged sword. Appropriate levels of autophagy can promote the removal of abnormal proteins or damaged organelles, while hyperactivated autophagy can induce autophagic apoptosis. However, recent studies suggest that autophagic flux seems to be blocked after traumatic brain injury (TBI), which contributes to the apoptosis of brain cells. In this study, valproic acid (VPA), which was clinically used for epilepsy treatment, was used to treat TBI. The Morris water maze test, hematoxylin & eosin staining and Nissl staining were first conducted to confirm that VPA treatment had a therapeutic effect on mice after TBI. Western blotting, enzyme-linked immunosorbent assay and immunofluorescence staining were then performed to reveal that VPA treatment reversed TBI-induced blockade of autophagic flux, which was accompanied by a reduced inflammatory response. In addition, the variations in activation and phenotypic polarization of microglia were observed after VPA treatment. Nevertheless, the use of the autophagy inhibitor 3-methyladenine partially abolished VPA-induced neuroprotection and the regulation of microglial function after TBI, resulting in the deterioration of the central nervous system microenvironment and neurological function. Collectively, VPA treatment reversed the TBI-induced blockade of autophagic flux in the mouse brain cortex, subsequently inhibiting brain cell apoptosis and affecting microglial function to achieve the promotion of functional recovery in mice after TBI. Cover Image for this issue: doi: 10.1111/jnc.14755.

We are in a machine learning age where several predictive applications that are life dependent are made by machines and robotic devices that relies on ensemble decision making algorithms. These have attracted many researchers and led to the development of an algorithm that is based on the integration of EKF, RBF networks and AdaBoost as an ensemble model to improve prediction accuracy. Firstly, EKF is used to optimize the slow training speed and improve the efficiency of the RBF network training parameters. Secondly, AdaBoost is applied to generate and combine RBFN-EKF weak predictors to form a strong predictor. Breast cancer survivability and diabetes diagnostic datasets used were obtained from the UCI repository. Results are presented on the proposed model as applied to Breast cancer survivability and Diabetes diagnostic predictive problems. The model outputs an accuracy of 96% when EKF-RBFN is applied as a base classifier compare to 94% when Decision Stump is applied and AdaBoost as an ensemble technique in both examples. The output accuracy of ensemble AdaBoostM1-Random Forest and standalone Random Forest models is 97% in both cases. The study has gone some way towards enhancing our knowledge and improving the prediction accuracy through the amalgamation of EKF, RBFN and AdaBoost algorithms as an ensemble model.

The success of automated lip reading has been constrained by the inability to distinguish between homopheme words, which are words that have different characters and produce the same lip movements (e.g. "time" and "some"), despite being intrinsically different. One word can often have different phonemes (units of sound) producing exactly the same viseme or visual equivalent of a phoneme for a unit of sound. Through the use of a Long-Short Term Memory Network with word embeddings, we can distinguish between homopheme words or words that produce identical lip movements. The neural network architecture achieved a character accuracy rate of 77.1% and a word accuracy rate of 72.2%.

In this paper, a new version of the Supervised t- Stochastic Neighbor Embedding (S-tSNE) algorithm is proposed which introduces the use of a dissimilarity measure related to class information. The proposed S-tSNE can be applied in any high dimensional dataset for visualization or as a feature extraction for classification problems. In this study, the S-tSNE is applied to three datasets MNIST, Chest x-ray, and SEER Breast Cancer. The two-dimensional data generated by the S-tSNE showed better visualization and an improvement in terms of classification accuracy in comparison to the original t- Stochastic Neighbor Embedding(t-SNE) method. The results from k-nearest neighbors (k-NN) classification model which used the lower dimension space generated by the new S-tSNE method showed more than 20% improvement on average in accuracy in all the three datasets compared with the t-SNE method. In addition, the classification accuracy using the S-tSNE for feature extraction was even higher than classification accuracy obtained from the original high dimensional data.

In this paper, we demonstrate how an existing deep learning architecture for automatically lip reading individuals can be adapted it so that it can be made speaker independent, and by doing so, improved accuracies can be achieved on a variety of different speakers. The architecture itself is multi-layered consisting of a convolutional neural network, but if we are to apply an initial edge detection-based stage to pre-process the image inputs so that only the contours are required, the architecture can be made to be less speaker favourable. The neural network architecture achieves good accuracy rates when trained and tested on some of the same speakers in the ”overlapped speakers” phase of simulations, where word error rates of just 1.3% and 0.4% are achieved when applied to two individual speakers respectively, as well as character error rates of 0.6% and 0.3%. The ”unseen speakers” phase fails to achieve as good an accuracy, with greater recorded word error rates of 20.6% and 17.0% when tested on the two speakers with character error rates of 11.5% and 8.3%. The variation in size and colour of different people’s lips will result in different outputs at the convolution layer of a convolutional neural network as the output depends on the pixel intensity of the red, green and blue channels of an input image so a convolutional neural network will naturally favour the observations of the individual whom the network was tested on. This paper proposes an initial ”contour mapping stage” which makes all inputs uniform so that the system can be speaker independent. Keywords: Lip Reading, Speech Recognition, Deep Learning, Facial Landmarks, Convolutional Neural Networks, Recurrent Neural Networks, Edge Detection, Contour Mapping

Compared to the traditional machine reading comprehension (MRC) with limitation to the information in a passage, knowledge-driven MRC tasks aim to enable models to answer the question according to text and related commonsense knowledge. Although pre-trained Transformer-based language models (TrLMs) such as BERT and Roberta, have shown powerful performance in MRC, external knowledge such as unspoken commonsense and world knowledge still cannot be used and explained explicitly. In this work, we present three simple yet effective injection methods integrated into the structure of TrLMs to fine-tune downstream knowledge-driven MRC tasks with off-the-shelf commonsense representations. Moreover, we introduce a mask mechanism for a token-level multi-hop relationship searching to filter external knowledge. We have conducted extensive experiments on DREAM and CosmosQA, two prevalent knowledge-driven datasets. Experimental results indicate that the incremental TrLMs have outperformed the baseline systems by 1%-4.1% with a fewer computational cost. Further analysis shows the effectiveness of the proposed methods and the robustness of the incremental model in the case of an incomplete training set.

Perfluorooctane sulfonate (PFOS) is a persistent organic pollutant that has been widely detected in the environment and has caused growing international concern. The liver is the main target organ of PFOS exposure. Animal experiments have shown that PFOS exposure can increase the risk of liver tumorigenesis. However, whether PFOS can accelerate liver tumor progression is still unclear. In this study, transgenic zebrafish Tg(fabp10:rtTA2s-M2; TRE2:EGFP-KRASG12V), a hepatocellular carcinoma (HCC) model that can cause liver tumorigenesis by doxycycline (DOX) induction, was used to investigate the effect of PFOS exposure in HCC progression. The male krasV12 transgenic zebrafish were exposed to 20 mg/L DOX, 500 μg/L PFOS or combined 20 mg/L DOX and 500 μg/L PFOS for 10 d. The results showed that co-treated with PFOS and DOX caused oncogenic Kras-induced liver enlargement, increased the percentages of zebrafish with HCC, and aggravated metabolic reprogramming of liver. To the best of our knowledge, this study for the first proved that PFOS could promote liver tumor progression. Decreased vitamin D level and increased fatty acid intake caused by PFOS might be responsible for the tumor-promoting effects. The results suggest that attention should be paid to the tumor-promoting effects of PFOS when assessing its environmental health risks, and these findings provide new insights into the toxicity of PFOS.

Species identification of human and animal blood is of critical importance in the areas of custom inspection, forensic science, wildlife preservation, and veterinary purpose. In this study, the combination of Raman spectroscopy and a recurrent neural network (RNN) is proposed for the discrimination of 20 kinds of blood species including human, poultry, wildlife, and experimental animals. The chemometric multi-classification model based on RNN was established and optimized by hyperparameter tuning and structure selection. The performance scores of the bidirectional RNN model with GRU for 20 kinds of species are as follows: accuracy 97.7%, precision 97.8%, recall 97.8% and F1-score 97.7%. The model resistant to wavenumber drift and cross-instrumental model were also studied for practical application purpose using a subset of Raman spectra by both commercial and laboratory-built Raman spectrometers. The evaluation shows an accuracy of 98.2%. These results indicate that our approach has great potential for blood species identification in real application scenarios.

The blood-spinal cord barrier plays a vital role in recovery after spinal cord injury. The neurovascular unit concept emphasizes the relationship between nerves and vessels in the brain, while the effect of the blood-spinal cord barrier on the neurovascular unit is rarely reported in spinal cord injury studies. Mouse models of spinal cord injury were established by heavy object impact and then immediately injected with platelet-derived growth factor (80 μg/kg) at the injury site. Our results showed that after platelet-derived growth factor administration, spinal cord injury, neuronal apoptosis, and blood-spinal cord barrier permeability were reduced, excessive astrocyte proliferation and the autophagy-related apoptosis signaling pathway were inhibited, collagen synthesis was increased, and mouse locomotor function was improved. In vitro, human umbilical vein endothelial cells were established by exposure to 200 μM H2O2. At 2 hours prior to injury, in vitro cell models were treated with 5 ng/mL platelet-derived growth factor. Our results showed that expression of blood-spinal cord barrier-related proteins, including Occludin, Claudin 5, and β-catenin, was significantly decreased and autophagy was significantly reduced. Additionally, the protective effects of platelet-derived growth factor could be reversed by intraperitoneal injection of 80 mg/kg chloroquine, an autophagy inhibitor, for 3 successive days prior to spinal cord injury. Our findings suggest that platelet-derived growth factor can promote endothelial cell repair by regulating autophagy, improve the function of the blood-spinal cord barrier, and promote the recovery of locomotor function post-spinal cord injury. Approval for animal experiments was obtained from the Animal Ethics Committee, Wenzhou Medical University, China (approval No. wydw2018-0043) in July 2018.

At present, the intelligent algorithms of multi-UAV task decision-making have been suffering some major issues, such as, slow learning speed and poor generalisation capability, and these issues hav...

Hydrodynamic characteristics of spawning grounds are important factors affecting the spawning of four major Chinese carps (Mylopharyngodon piceus, Ctenopharyngodon idella, Hypophthalmichthys molitrix, and Aristichthys nobilis). To investigate the relationship between the preferred hydrodynamic characteristics of spawning sites and the response of fish spawning behavior, we monitored the flow field of spawning sites during ecological operation of the Three Gorges Dam (i.e., man-made flood regulation) in 2014 and 2015. We used the data to explore the correlation between vorticity changes in spawning grounds and the spawning amount. Pearson correlation coefficients of the average vorticity in all cross-sections of the Yidu spawning ground and spawning amount in 2014 and 2015 were 0.730 and 0.822, respectively, indicating a significant positive correlation between vorticity and spawning activity. In some specific regions, this correlation was even stronger (Pearson correlations of the regional vorticity and egg production were &gt, 0.95). To further corroborate and analyze the relationship between these specific regions and the distribution of broodstock during the breeding season, an ultrasonic telemetry test of broodstock was conducted in the Yidu spawning ground in 2016. The results showed that the broodstocks were concentrated in the reach near the Quantong Pier (~76 km from the Three Gorges Dam). These regions were consistent with areas of river characterized by highly correlated vorticity and egg production levels, suggesting that these regions are areas preferred by four major Chinese carps for spawning.

Severe acute pancreatitis (SAP) is a digestive system disease that is associated with a range of complications including intestinal dysfunction. In this study, we determined that the chalcone compound, isoliquiritigenin (ISL), reduces pancreatic and intestinal injury in a mouse model of SAP. These effects were achieved by suppressing oxidative stress and the inflammatory responses to SAP. This was evidenced by a reduction in histological score, and malondialdehyde (MDA), interleukin (IL)-6, tumor necrosis factor (TNF)-α and cleaved-caspase-3 (c-caspase-3) protein along with an increase in Nrf2, hemeoxygenase-1 (HO-1), quinone oxidoreductase 1 (NQO1), and superoxide dismutase (SOD). We then used Nrf2-/- mice to test the protective effect of Nrf2 during ISL treatment of SAP. Our results indicated that Nrf2-/- mice had greater pancreatic injury and intestinal dysfunction than wild-type mice. They also had reduced adherens junctions (P120-catenin) and tight junctions (occludin), and increased activated nuclear factor-κB (NF-κB) protein. In Nrf2-/- mice, ISL was less effective at these functions than in the WT mice. In conclusion, this study demonstrated that ISL exerts its protective effects against oxidative stress and inflammatory injury after SAP via regulation of the Nrf2/NF-κB pathway. It also showed that the efficacy of ISL in repairing the intestinal barrier damage caused by SAP is closely related to the Nrf2 protein. Our findings demonstrated that Nrf2 is an important protective factor against SAP-induced injuries in the pancreas and intestines.

The construction of the Three Gorges Dam, along with other development in the Yangtze River basin, has had profound consequences for the river’s flow and sediment regime. This has had major impacts on the geomorphology and ecology of the river downstream of the dam, with related impacts on biodiversity, including fish populations, livelihoods, and water security in the middle and lower Yangtze. Changes to fish populations have included a fall of around 90% in the total number of fish fry for the four economically-important Chinese carp species, caused at least in part by alterations in the flow regime. In response, there has been increased research into the significance of flow regimes for Chinese carp, as well as other aspects of river health. A partnership between the Chinese Government, the dam operator, scientists, and conservationists has led to pilot environmental flow releases over a five-year period in an attempt to mitigate some of these impacts. Subsequent monitoring has shown that numbers of fish fry are increasing from the low they had fallen to in 2008. Drawing on lessons from the pilot environmental flow releases, in October 2015 the official regulations that govern operations of the Three Gorges Dam were amended to incorporate additional objectives, including incorporating environmental flow releases as part of the routine operation of the dam. This paper describes the processes that led to the environmental flow program from Three Gorges, a review of monitoring data collected during the pilot environmental flow releases, the subsequent amendment of the dam operating rules, and prospects for expanding environmental flow implementation in the Yangtze River in coming years.

Traumatic brain injury (TBI) is a serious public health and medical problem worldwide. Oxidative stress plays a vital role in the pathogenesis of TBI. Nuclear factor erythroid 2-related factor 2 (Nrf2), an important factor in the cellular defense against oxidative stress, is activated following TBI. In this study, the protective effects of Isoliquiritigenin (ILG), a promising antioxidant stress drug, was evaluated as a protective agent against TBI. In a mouse model of controlled cortical impact Injury, we found that the ILG administration reduced the Garcia neuroscore, injury histopathology, brain water content, cerebral vascular permeability, the expression of cleaved caspase3, aquaporin-4, glial fibrillary acidic protein and the increased the expression of neurofilament light chain protein, indicating the protective effects against TBI in vivo. ILG treatment after TBI also restored the oxidative stress and promoted the Nrf2 protein transfer from the cytoplasm to the nucleus. We then used Nrf2-/- mice to test the protective effect of Nrf2 during ILG treatment of TBI. Our findings indicated that Nrf2-/- mice had greater brain injury and oxidative stress than wild-type (WT) mice and ILG was less effective at inhibiting oxidative stress and repairing the brain injury than in the WT mice. In vitro studies in SY5Y cells under oxygen glucose deprivation/re-oxygenation stimulation yielded results that were consistent with those obtained in vivo showing that ILG promotes Nrf2 protein transfer from the cytoplasm to the nucleus. Taken together, our findings demonstrate that Nrf2 is an important protective factor against TBI-induced injuries, which indicates that the protective effects of ILG are mediated by inhibiting oxidative stress after TBI via a mechanism that involves the promotion of Nrf2 protein transfer from the cytoplasm to the nucleus.

A novel method for multi-plane imaging in digital holography is proposed: holograms of objects located at various places are recorded and then reconstructed simultaneously through one-step Fresnel diffraction using the quadratic distorted phase factor (QDPF). The theory of one-step Fresnel diffraction with the QDPF is deduced and experimental results support the theoretical investigation.

This paper presents the empirical comparison of boosting implementation by reweighting and resampling methods. The goal of this paper is to determine which of the two methods performs better. In the study, we used four algorithms namely: Decision Stump, Neural Network, Random Forest and Support Vector Machine as base classifiers and AdaBoost as a technique to develop various ensemble models. We applied 10-fold cross validation method in measuring and evaluating the performance metrics of the models. The results show that in both methods the average of the correctly classified and incorrectly classified are relatively the same. However, average values of the RMSE in both methods are insignificantly different. The results further show that the two methods are independent of the datasets and the base classier used. Additionally, we found that the complexity of the chosen ensemble technique and boosting method does not necessarily lead to better performance.

According to the inherent requirements of education for talents' knowledge, quality and comprehensive ability and the major training goals of optoelectronics information science and engineering, in order to enhance the undergraduates' comprehensive practical ability and consciousness of innovation, we carried out the reforms of teaching method and teaching mode, which took the training programs of innovation and entrepreneurship for undergraduates, extracurricular academic research fund, "Chun-Tsung Scholar" program or research projects of their tutors as the guidance, and took the all levels of relevant discipline competitions as the promotion. And the training mainline of engineering innovation talents as "undergraduate's tutorial system →innovative training program or tutor’s research project →academic competition →graduation projects (thesis)" was constructed stage by stage by combining the undergraduates' graduation projects and their participated academic competition into one for improving the quality of the graduation projects (thesis). The practical results of the last several years illuminate that the proposed training model can effectively stimulate the students' awareness of autonomous learning, enhance their comprehensive ability of analyzing and solving problems and improve their ability of engineering practice and innovation as well as their teamwork spirit.

The effects of surgically implanted dummy ultrasonic transmitters on growth, survival and transmitter retention of bighead carp Hypophthalmichthys nobilis were investigated from October to November 2010. Before the surgery, 162 bighead carps were divided into three groups: 1) control group: handling but no tagging; 2) sham group: surgical procedure without implantation of dummy ultrasonic transmitter; 3) surgery group: surgical implantation of dummy ultrasonic transmitters. Growth, survival, incision healing, suture loss and transmitter retention of bighead carp were recorded and analyzed respectively at 14, 28 and 56 days after surgery. The results showed that surgical implantation of dummy ultrasonic transmitters had a significantly negative influence on growth of bighead carp, however, the negative influence was negligible 28 days after surgery. No fish died during the experimental period. All incisions had completely healed 56 days after surgery, but sutures remained in 21 fish, and the suture loss rate was 70.8 %. All fish with suture retention had suffered light infection over the incision sites. No dummy ultrasonic transmitters were expelled 56 days after the surgery. Except for three transmitters, all the remaining transmitters were completely encapsulated in a fibrous capsule and adhered to the body wall, intestinal tract or viscera 56 days after surgery. The above results suggested that telemetry research could be conducted on bighead carp within 56 days after surgical implantation of ultrasonic transmitters.

The Nu River Project comprised of 13-dam cascade is one of the biggest hydropower development projects yet conceived in China. However, hydropower development will inevitably result in the local ecological environment and biological material impact, particularly the effects on fish. Glyptothorax zainaensis is weak in swimming with suction cup and is extremely sensitive in the hydrological characteristics and water quality changes, so that it can be considered as a water quality monitor of indicator organism. Currently, with the need for conservation, fishery restoration and so on, there is a great demand for genetic information on Glyptothorax zainaensis. Fourty microsatellite loci were isolated, characterized and evaluated from Glyptothorax zainaensis. A total of 91 different alleles were observed in the 303 wild Glyptothorax zainaensis, and the percentage of polymorphic bands ranged from 41.8% to 66.7%, with an average of 90.1%. The genetic diversity value was significantly huge with the expected heterozygosity ranging from 0.578 to 0.706, the observed heterozygosity value ranging from 0.923 to 0.980 and PIC value ranging from 0.418 to 0.644, which showed that the wild Glyptothorax zainaensis population had a high level of genetic diversity. Analysis of molecular variance (AMOVA) and pairwise Fst showed a significant level of genetic differentiation (P

The flight attitude control is the core part of the maneuvering process in air combats. Traditional flight attitude control methods have high computational complexity, low flexibility and poor ability to learn sequential feature. This paper proposes a flight attitude control model based on long short term memory network, which utilizes its special gates structure to memorize historical information, and acquire the variation law of the attitude control variable from the time sequential data including the battlefield situation and flight parameters automatically. Moreover, the basic framework and training methods of the model are also introduced, and the influence caused by various LSTM network parameters is deeply discussed. The experiment results show that the proposed model has better prediction accuracy and convergence performance than the traditional recurrent neural network.

An improved multi-plane imaging fusion algorithm for 3-D digital holography is presented. Multiple reconstructed images at different distances from single off-axis Fresnel digital hologram are fused using discrete cosine transform (DCT) with different blocks. The zero order term is removed using a Laplacian operator while the conjugate image and dispersed images of defocused objects are eliminated by selecting a simple rectangular window before image fusion. Compared with those fused images obtained directly from the reconstructed images, the fused images qualities are greatly improved and all of the objects will be clearly in focus more quickly than discrete wavelet (DWT) based fusion method.

Parkinson's disease (PD) is a progressive neurodegenerative disorder with complicated pathophysiologic mechanisms. Endoplasmic reticulum (ER) stress appears to play a critical role in the progression of PD. We demonstrated that basic fibroblast growth factor (bFGF), as a neurotropic factor, inhibited ER stress-induced neuronal cell apoptosis and that 6-hydroxydopamine (6-OHDA)-induced ER stress was involved in the progression of PD in rats. bFGF administration improved motor function recovery, increased tyrosine hydroxylase (TH)-positive neuron survival, and upregulated the levels of neurotransmitters in PD rats. The 6-OHDA-induced ER stress response proteins were inhibited by bFGF treatment. Meanwhile, bFGF also increased expression of TH. The administration of bFGF activated the downstream signals PI3K/Akt and Erk1/2 in vivo and in vitro. Inhibition of the PI3K/Akt and Erk1/2 pathways by specific inhibitors partially reduced the protective effect of bFGF. This study provides new insight towards bFGF translational drug development for PD involving the regulation of ER stress.

A new approach of high accuracy displacement field measurement in long-distance by digital image correlation is proposed. The long-distance measurement accuracy attained 1 micron with digital image correlation technique of 0.001 pixel bilinear interpolation. Integrated software package is exploited with language of c++ and Matlab under Windows system . It combined with optics hardware to constitute a multi-function instrument for displacement field measurement. This system have features of high accuracy, real-time and non-contact measurement with computer control and is especially effective in flexivity measurement of bridge and cloverleaf junction. It also fit for dynamic displacement field measurement when a high speed image card was used. This technology extend digital image correlation method's application field.

Many small online retailers and new entrants to the online retail sector are keen to practice data mining and consumer-centric marketing in their businesses yet technically lack the necessary knowledge and expertise to do so. In this article a case study of using data mining techniques in customer-centric business intelligence for an online retailer is presented. The main purpose of this analysis is to help the business better understand its customers and therefore conduct customer-centric marketing more effectively. On the basis of the Recency, Frequency, and Monetary model, customers of the business have been segmented into various meaningful groups using the k-means clustering algorithm and decision tree induction, and the main characteristics of the consumers in each segment have been clearly identified. Accordingly a set of recommendations is further provided to the business on consumer-centric marketing. SAS Enterprise Guide and SAS Enterprise Miner are used in the present study.