Toxic Comment Classification Based on Bidirectional Gated Recurrent Unit and Convolutional Neural Network

For English toxic comment classification, this paper presents the model that combines Bi-GRU and CNN optimized by global average pooling (BG-GCNN) based on the bidirectional gated recurrent unit (Bi-GRU) and global pooling optimized convolution neural network (CNN) . The model treats each type of toxic comment as a binary classification. First, Bi-GRU is used to extract the time-series features of the comment and then the dimensionality is reduced through global pooling optimized convolution neural network. Finally, the classification result is output by Sigmoid function. Comparative experiments show the BG-GCNN model has a better classification effect than Text-CNN, LSTM, Bi-GRU, and other models. The Macro-F1 value of the toxic comment dataset on the Kaggle competition platform is 0.62. The F1 values of the three toxic label classification results (toxic, obscene, and insult label) are 0.81, 0.84, and 0.74, respectively, which are the highest values in the comparative experiment.

An Enhanced Sliding Mode Speed Control for Induction Motor Drives

In this paper, an enhanced Integral Sliding Mode Control (ISMC) for mechanical speed of an Induction Motor (IM) is presented and experimentally validated. The design of the proposed controller has been done in the d-q synchronous reference frame and indirect Field Oriented Control (FOC). Global asymptotic speed tracking in the presence of model uncertainties and load torque variations has been guaranteed by using an enhanced ISMC surface. Moreover, this controller provides a faster speed convergence rate compared to the conventional ISMC and the Proportional Integral methods, and it eliminates the steady-state error. Furthermore, the chattering phenomenon is reduced by using a switching sigmoid function. The stability of the proposed controller under parameter uncertainties and load disturbances has been provided by using the Lyapunov stability theory. Finally, the performance of this control method is verified through numerical simulations and experimental tests, getting fast dynamics and good robustness for IM drives.

Short communication: A new tool to define multiscale bedform characteristics from bed elevation data

Systematic identification and characterization of bedforms from bathymetric data are crucial in many studies focused on fluvial processes. Automated and accurate processing of bed elevation data is challenging where dune fields are complex, irregular and, especially, where multiple scales co-exist. Here, we introduce a new tool to quantify dune properties from bathymetric data representing multiple dune scales. A first step in the procedure is to decompose the bathymetric data based on a LOESS algorithm. Steep dune lee side slopes are accounted for by implementing objective breaks in the algorithm, accounting for discontinuities in the bed level profiles, often occurring at the toe of the lee side slope of dunes. The steep lee slopes are then approximated by fitting a sigmoid function. Following the decomposition of the bathymetric data, bedforms are identified based on zero-crossing, and the relevant properties are calculated. The approach to decompose bedforms adopted in the presented tool is particularly applicable where secondary dunes are large and thus filtering could easily lead to undesired smoothing of the primary morphology. Application of the tool to two bathymetric maps demonstrates that the decomposition and identification are successful, as the lee side slopes are better preserved.

Recommendation algorithm based on attributed multiplex heterogeneous network

In the field of deep learning, the processing of large network models on billions or even tens of billions of nodes and numerous edge types is still flawed, and the accuracy of recommendations is greatly compromised when large network embeddings are applied to recommendation systems. To solve the problem of inaccurate recommendations caused by processing deficiencies in large networks, this paper combines the attributed multiplex heterogeneous network with the attention mechanism that introduces the softsign and sigmoid function characteristics and derives a new framework SSN_GATNE-T (S represents the softsign function, SN represents the attention mechanism introduced by the Softsign function, and GATNE-T represents the transductive embeddings learning for attribute multiple heterogeneous networks). The attributed multiplex heterogeneous network can help obtain more user-item information with more attributes. No matter how many nodes and types are included in the model, our model can handle it well, and the improved attention mechanism can help annotations to obtain more useful information via a combination of the two. This can help to mine more potential information to improve the recommendation effect; in addition, the application of the softsign function in the fully connected layer of the model can better reduce the loss of potential user information, which can be used for accurate recommendation by the model. Using the Adam optimizer to optimize the model can not only make our model converge faster, but it is also very helpful for model tuning. The proposed framework SSN_GATNE-T was tested for two different types of datasets, Amazon and YouTube, using three evaluation indices, ROC-AUC (receiver operating characteristic-area under curve), PR-AUC (precision recall-area under curve) and F1 (F1-score), and found that SSN_GATNE-T improved on all three evaluation indices compared to the mainstream recommendation models currently in existence. This not only demonstrates that the framework can deal well with the shortcomings of obtaining accurate interaction information due to the presence of a large number of nodes and edge types of the embedding of large network models, but also demonstrates the effectiveness of addressing the shortcomings of large networks to improve recommendation performance. In addition, the model is also a good solution to the cold start problem.

Sliding Dispersion Entropy-Based Fault State Detection for Diaphragm Pump Parts

Aiming at the disadvantages of low trend, poor characterization performance, and poor anti-noise performance of traditional degradation features such as dispersion entropy (DE), a fault detection method based on sliding dispersion entropy (SDE) is proposed. Firstly, a sliding window is added to the signal before extracting the DE feature, and the root mean square of the signal inside the sliding window is used to replace the signal in the window to realize down sampling, which enhances the trend of DE. Secondly, the hyperbolic tangent sigmoid function (TANSIG) is introduced to map the signals to different categories when extracting the DE feature, which is more in line with the signal distribution of mechanical parts and the monotonicity of the degradation feature is improved. For noisy signal, the introduction of locally weighted scatterplot smoothing (LOWESS) can remove the burrs and fluctuations of the SDE curve, and the anti-noise performance of SDE is improved. Finally, the SDE state warning line is constructed based on the 2σ criterion, which can determine the fault warning point in time and effectively. The state detection results of bearing and check valve show that the proposed SDE improves the trend, monotonicity, and robustness of the state tracking curve, and provides a new method for fault state detection of mechanical parts.

Computer Multimedia Art Pattern Creation Based on P-Filling Algorithm

Art education is an important part of quality education in China. It undertakes the important responsibility of cultivating students' aesthetic ability, art appreciation ability, art perception ability, and expression ability. In junior middle school art teaching, the skill of pattern creation is an ability that needs to be focused on cultivating students. The computer multimedia art patterns are currently created mainly relying on manual drawing. A computer multimedia art pattern production model is proposed based on the P-filling algorithm in this paper. After using the P-filling algorithm to quickly generate a large number of fake images and retrieve the code of the most recent damaged image, the code estimates the missing content by generating a model. On this basis, the semantic P-filling algorithm and the perceptual P-filling algorithm are combined, and the unsaturated region is enlarged by improving the activation function sigmoid function, which solves the problem that the gradient is easy to disappear. Experimental results show that, through this production model, users can produce a large number of computer multimedia art patterns with hand-drawn style features through very little interaction and parameter control.

Empirical method for modeling the percent depth dose curves of electron beam in radiation therapy

Introduction: This study presents an empirical method to model the electron beam percent depth dose curve (PDD) using the primary and tail functions in radiation therapy. The modeling parameters N and n can be used to derive the depth relative stopping power of the electron energy in radiation therapy. Methods and Materials: The electrons PDD curves were modeled with the primary-tail function in this study. The primary function included exponential function and main parameters of N, µ while the tail function was composed by a sigmoid function with the main parameter of n. The PDD for five electron energies were modeled by the primary and tail function by adjusting the parameters of N, µ and n. The R50 and Rp can be derived from the modeled straight line of 80% to 20% region of PDD. The same electron energy with different cone sizes was also modeled with the primary-tail function. The stopping power for different electron energies at different depths can also be derived from the parameters of N, µ and n. Percent ionization depth curve can then be derived from the percent depth dose by dividing its depth relevant stopping power for comparing with the original water phantom measurement. Results: The main parameters N, n increase, but µ decreases in primary-tail function when electron energy increased. The relationship of parameters n, N and LN(-µ) with electron energy are n = 31.667 E0 - 88, N = 0.9975 E0 - 2.8535, LN(-µ) = -0.1355 E0 - 6.0986, respectively. Stopping power of different electron energy can be derived from n and N with the equation: stopping power = (−0.042 ln N E 0 + 1.072)e(−n−E0·5·10−5+0.0381·d), where d is the depth in water. Percent depth dose was derived from the percent reading curve by multiplying the stopping power relevant to the depth in water at certain electron energy. Conclusion: The PDD of electrons at different energies and field sizes can be modeled with an empirical model to deal with the stopping power calculation. The primary-tail equation provides a uncomplicated solution than a pencil beam or other numerical algorism for investigators to research the behavior of electron beam in radiation therapy.