In the context of global and regional hydrometeorological environmental changes, extreme weather events have become increasingly common, resulting in heavy precipitation and flooding. This has led to an increased risk...
详细信息
In the context of global and regional hydrometeorological environmental changes, extreme weather events have become increasingly common, resulting in heavy precipitation and flooding. This has led to an increased risk of rainfall-runoff pollution. This study investigated the evolution of water quality in Dahuofang Reservoir (DHFR) under the influence of rainfall-runoff, utilising water quality monitoring data from the DHFR in Northeast China. Furthermore, an encoder-decoder model based on Long Short-Term Memory (LSTM-ED) was constructed to predict the water quality indicators, and the response of these indicators to different levels of rainfall-runoff was subsequently explored. The findings demonstrate that elevated precipitation levels led to an increase in pH and the concentrations of the permanganate index (CODMn) and total phosphorus (TP) in the DHFR, while the concentrations of dissolved oxygen (DO) and total nitrogen (TN) exhibited a decline. The LSTM-ED model constructed in this study was effective in predicting the changes in water quality indicators in DHFR. It was observed that as the characteristic variables affecting the water quality indicators increased gradually, the concentration of each water quality indicator exhibited an increasing trend. Notably, the amplitude of the pH increase gradually diminished, while the amplitude of the TN and DO increases gradually increased. In particular, our findings indicated that the concentrations of TN and CODMn will reach the standards for inferior Class V and Class III-IV surface water, respectively, when the characteristic variables affecting TN and CODMn concentrations reached their historical maximum levels.
Atmospheric PM2.5 is a major pollutant impacting on human health and the environment. Based on traditional neural networks, we construct three prediction models: BP, Stack GRU, and encoder-decoder. We use Tianjin'...
详细信息
ISBN:
(纸本)9781538650592
Atmospheric PM2.5 is a major pollutant impacting on human health and the environment. Based on traditional neural networks, we construct three prediction models: BP, Stack GRU, and encoder-decoder. We use Tianjin's continuous 171-day hourly PM2.5 concentration, air quality index, and meteorological data to train and test the model and evaluate the prediction performance of the model. In the experiment, using the meteorological factors, pollutant factors, seasonal factors, and PM2.5 concentrations as inputs, the PM2.5 concentration of every hour of the next day is predicted. The experimental result shows that when using PM2.5 concentration data for 3 days per hour to predict PM2.5 per hour, continuous forecasting for 43 days, the PM2.5 concentration value predicted by the encoder-decoder model is not significantly different from the value of PM2.5 issued by Tianjin local authorities, and the root mean square error (RMSE) is 43.17. With the same input data, the prediction result of encoder-decoder model is better than BP neural network and GRU prediction model, which shows that encoder-decoder model has better adaptability in predicting PM2.5 concentration than BP neural network and GRU model.
Chinese couplet consists of two syntactically symmetric sentences which are concise but meaningful and equal in length. Creating Chinese couplets is a challenging task even in the artificial intelligence field. Couple...
详细信息
Chinese couplet consists of two syntactically symmetric sentences which are concise but meaningful and equal in length. Creating Chinese couplets is a challenging task even in the artificial intelligence field. Couplet automatic generation approaches based on maximum likelihood estimation (MLE) have been studied. However, MLE-based methods are limited by exposure bias, and the token-level probability maximizing generation makes the model ignore the global control of sentences, while couplets also have strict requirements at the sentence level between two sentences, such as semantic and sentence pause. In this paper, we propose a GAN-based model to fill this gap, named CoupGAN. Specifically, we introduce a discriminative model to evaluate the global-level quality of generated sentences. Meanwhile, to deal with discrete text data, CoupGAN is trained by a policy gradient-based method. In detail, this method rewards the generator to mitigate exposure bias, which could encourage the model to generate more realistic samples. Experiments among large Chinese couplet corpora datasets demonstrate that our proposed model is superior to existing methods in various evaluation metrics.
Methane is one of the most dangerous gases produced in the process of coal mining. Because of its flammable and explosive characteristics, it has seriously threatened the life and property safety of coal miners. As a ...
详细信息
Methane is one of the most dangerous gases produced in the process of coal mining. Because of its flammable and explosive characteristics, it has seriously threatened the life and property safety of coal miners. As a result, accurate and real-time gas concentration forecasting is becoming a crucial but challenging issue for reducing methane risks and accidents. To further improve the efficiency and accuracy of methane concentration forecasting, this paper proposes a graph convolutional encoder-decoder (GCN-ED) network, which can train and infer all the sensors of a coal face as a unified entity. The proposed GCN-ED is composed of the GCN module and the ED module with a parallel structure. The GCN module constructs a priori graph structure through the adjacency relation between sensors in reality and uses a learnable self-adaptive dependency matrix to precisely capture the hidden spatial dependency in the data. The ED module is used to learn complex temporal features with LSTM cells and generate multi-step results of the gas concentration prediction. Experiments are conducted on real coal mine datasets, whose results demonstrate that the GCN-ED achieves the better performance than various state-of-the-art solutions and largely improves the efficiencies of training processes.
The concentration of fine particulate matter (PM2.5), which represents inhalable particles with diameters of 2.5 micrometers and smaller, is a vital air quality index. Such particles can penetrate deep into the human ...
详细信息
The concentration of fine particulate matter (PM2.5), which represents inhalable particles with diameters of 2.5 micrometers and smaller, is a vital air quality index. Such particles can penetrate deep into the human lungs and severely affect human health. This paper studies accurate PM2.5 prediction, which can potentially contribute to reducing or avoiding the negative consequences. Our approach's novelty is to utilize the genetic algorithm (GA) and an encoder-decoder (E-D) model for PM2.5 prediction. The GA benefits feature selection and remove outliers to enhance the prediction accuracy. The encoder-decoder model with long short-term memory (LSTM), which relaxes the restrictions between the input and output of the model, can be used to effectively predict the PM2.5 concentration. We evaluate the proposed model on air quality datasets from Hanoi and Taiwan. The evaluation results show that our model achieves excellent performance. By merely using the E-D model, we can obtain more accurate (up to 53.7%) predictions than those of previous works. Moreover, the GA in our model has the advantage of obtaining the optimal feature combination for predicting the PM2.5 concentration. By combining the GA-based feature selection algorithm and the E-D model, our proposed approach further improves the accuracy by at least 13.7%.
As a common and high-risk type of disease,heart disease seriously threatens people’s *** the same time,in the era of the Internet of Thing(IoT),smart medical device has strong practical significance for medical worke...
详细信息
As a common and high-risk type of disease,heart disease seriously threatens people’s *** the same time,in the era of the Internet of Thing(IoT),smart medical device has strong practical significance for medical workers and patients because of its ability to assist in the diagnosis of ***,the research of real-time diagnosis and classification algorithms for arrhythmia can help to improve the diagnostic efficiency of *** this paper,we design an automatic arrhythmia classification algorithm model based on Convolutional Neural Network(CNN)and encoder-decoder *** model uses Long Short-Term Memory(LSTM)to consider the influence of time series features on classification ***,it is trained and tested by the MIT-BIH arrhythmia ***,Generative Adversarial Networks(GAN)is adopted as a method of data equalization for solving data imbalance *** simulation results show that for the inter-patient arrhythmia classification,the hybrid model combining CNN and encoder-decoder model has the best classification accuracy,of which the accuracy can reach 94.05%.Especially,it has a better advantage for the classification effect of supraventricular ectopic beats(class S)and fusion beats(class F).
Geospatial studies must address spatial data quality, especially in data-driven research. An essential concern is how to fill spatial data gaps (missing data), such as for cartographic polylines. Recent advances in de...
详细信息
Geospatial studies must address spatial data quality, especially in data-driven research. An essential concern is how to fill spatial data gaps (missing data), such as for cartographic polylines. Recent advances in deep learning have shown promise in filling holes in images with semantically plausible and context-aware details. In this paper, we propose an effective framework for vector-structured polyline completion using a generative model. The model is trained to generate the contents of missing polylines of different sizes and shapes conditioned on the contexts. Specifically, the generator can compute the content of the entire polyline sample globally and produce a plausible prediction for local gaps. The proposed model was applied to contour data for validation. The experiments generated gaps of random sizes at random locations along with the polyline samples. Qualitative and quantitative evaluations show that our model can fill missing points with high perceptual quality and adaptively handle a range of gaps. In addition to the simulation experiment, two case studies with map vectorization and trajectory filling illustrate the application prospects of our model.
The multi-function radar (MFR) is capable of transmitting complex signals and achieving beam agility according to different tasks, thus leading to many challenges in the work mode recognition field. This paper, theref...
详细信息
ISBN:
(数字)9781665427920
ISBN:
(纸本)9781665427920
The multi-function radar (MFR) is capable of transmitting complex signals and achieving beam agility according to different tasks, thus leading to many challenges in the work mode recognition field. This paper, therefore, develops an encoder-decoder model based on the gated recurrent units (GRU) network to achieve work mode recognition. It involves the use of the encoder structure to extract the temporal features and the work mode transition regulations of the intercepted pulse group sequence while leveraging the decoder structure to decode the features and transition regulations as a part of the input for the next decoding process. Additionally, the label substitution (LS) method is utilized for improving the ability of the decoder structure to recognize work mode under non-ideal situations. Simulation results show that the proposed method is less sensitive in the presence of non-ideal situations and shares better performance than existing work mode recognition methods.
Slot filling task, which aims to predict the semantic slot labels for each specific word in word sequence, is one of the main tasks in Spoken Language Understanding (SLU). In this paper, we propose a variation of enco...
详细信息
ISBN:
(纸本)9783319773834;9783319773827
Slot filling task, which aims to predict the semantic slot labels for each specific word in word sequence, is one of the main tasks in Spoken Language Understanding (SLU). In this paper, we propose a variation of encoder-decoder model for sequence labelling. To better use the label dependency feature and prevent overfitting, we use Long Short Term Memory (LSTM) as encoder and Gated Recurrent Unit (GRU) as decoder. We also enhance the model by employing the attention mechanism with attention window as a novel feature, which considers the particularity in slot filling task that each target label corresponds to the specific words and hidden units in the encoder. We test the proposed model using the standard ATIS corpus by adopting different size of attention window. The analysis of trends for the results using different attention window size has shown its application potential of attention window feature.
The standard evaluation metric of automatic speech recognition (ASR) is the word error rate (WER), which measures the dissimilarity between recognized word sequences and their ground truth. Many training algorithms de...
详细信息
ISBN:
(纸本)9781538646588
The standard evaluation metric of automatic speech recognition (ASR) is the word error rate (WER), which measures the dissimilarity between recognized word sequences and their ground truth. Many training algorithms designed to reduce sequence-level errors such as WER have been proposed for hidden Markov model (HMM)-based ASR, e.g., state-level minimum Bayes risk (sMBR). However, these approaches cannot be used directly for encoder-decoder model based end-to-end ASR, because the encoder-decoder model employs very different mechanisms from HMM-based approaches. In this paper, we propose a new method for optimizing the encoder-decoder model based on a sequence-level evaluation metric. Since the WER is not directly differentiable, we adopt a policy gradient objective function to train the encoder-decoder model, which enables us to minimize the expected WER of the model predictions. This training method employs the scoring of multiple hypotheses as in the decoding stage while usual cross entropy training uses only the ground truth. Therefore, we can expect it to improve the decoding results of the encoder-decoder model. We perform experiments using the Tedlium corpus to demonstrate the potential of our proposed method for improving the recognition performance of the encoder-decoder model.
暂无评论