The methods for RUL prediction of bearings are mainly based on the autoregressive strategies, among which the temporal convolutional network (TCN) has been recently developed and widely believed as ...the high-performance one. These methods generally suffer from the errors of prediction. In this paper, we newly design the Bayesian-optimization-based adversarial temporal convolutional network (AdTCN-BO), by embedding the TCN into the adversarial training framework as the generator. Within the framework, the discriminator is designed to continuously correct the output value of the generator in the training process, thus reducing the errors of prediction to a certain extent. Based on the AdTCN-BO, a novel RUL prediction approach for bearings is developed. An experiment verification is carried out to validate the effectiveness of the proposed approach, demonstrating that the AdTCN-BO framework is more accurate in contrast to the traditional data-driven methods of RUL prediction.
Power stations operating on PhotoVoltaic (PV) power generation are challenged by demand forecasting as PV power generation is random and intermittent. This paper presents an Efficient Shrinkage ...Temporal Convolutional Network (ESTCN) model, which combines the Temporal Convolutional Network (TCN) and an improved Deep Residual Shrinkage Network (DRSN) to forecast PV power output. First, the attention sub-network of the DRSN is improved, the fully connected layer of the sub-network is canceled, and feature extraction is done directly on the model features following global average pooling using a one-dimensional convolution to obtain cross-channel interaction and increased model efficiency. Next, an improved attention mechanism and adaptive soft thresholding are introduced into TCN to automatically determine the noise threshold to address the issue of information weight dispersion caused by redundant information in the input samples. By incorporating dilated causal convolution, attention module, soft thresholding, and residual connection, the ESTCN model is formed and is shown to enhance PV power output prediction with only a minimal increase in the number of parameters. The power station in Ningxia, China is adopted as a validation example with data taken from the year 2020. The fitting and prediction results of the ESTCN model are compared against the Convolutional Neural Network, Long Short-Term Memory, and TCN models. The ESTCN model yielded the following values of the evaluation metrics: Root Mean Square Error (RMSE) of 1.47 kW, Mean Absolute Error (MAE) of 0.79 kW, and coefficient of determination of 0.999. Compared to the TCN model, the prediction errors based on RMSE and MAE improved by 0.39 kW and 0.18 kW, respectively. Applying the ESTCN model to predict PV power generation of power stations in two regions in China, Ningxia and Xinjiang, shows the ESTCN model to be superior, scalable, and universal over other deep learning models.
•Study the problem of photovoltaic power using a deep learning method.•A new Efficient Shrinkage Temporal Convolutional Network model is presented.•Attention sub-network of DRSN is improved by cancelling the fully connected layer.•Feature extraction is done on the model features using a one-dimensional convolution.•Improved attention mechanism and adaptive soft thresholding are introduced into TCN.
•An efficient TCNet-Fusion model for MI-EEG classification is proposed.•1D convolutions are applied in temporal domain and channel-wise in order.•An image-like 2D representation is fed to the ...proposed model.•The model achieved 83.73 % accuracy in BCI Competition IV-2a.•The model achieved 94.41 % accuracy in High Gamma Dataset.
Motor imagery electroencephalography (MI-EEG) signals are generated when a person imagines a task without actually performing it. In recent studies, MI-EEG has been used in the rehabilitation process of paralyzed patients, therefore, decoding MI-EEG signals accurately is an important task, and it is difficult task due to the low signal-to-noise ratio and the variation of brain waves between subjects. Deep learning techniques such as the convolution neural network (CNN) have shown an impact in extracting meaningful features to improve the accuracy of classification. In this paper, we propose TCNet-Fusion, a fixed hyperparameter-based CNN model that utilizes multiple techniques, such as temporal convolutional networks (TCNs), separable convolution, depth-wise convolution, and the fusion of layers. This model outperforms other fixed hyperparameter-based CNN models while remaining similar to those that utilize variable hyperparameter networks, which are networks that change their hyperparameters based on each subject, resulting in higher accuracy than fixed networks. It also uses less memory than variable networks. The EEG signal undergoes two successive 1D convolutions, first along with the time domain, then channel-wise. Then, we obtain an image-like representation, which is fed to the main TCN. During experimentation, the model achieved a classification accuracy of 83.73 % on the four-class MI of the BCI Competition IV-2a dataset, and an accuracy of 94.41 % on the High Gamma Dataset.
Accurate and timely traffic forecasting is significant for intelligent transportation management. However, existing approaches model the temporal and spatial features of traffic flow inadequately. To ...address these limitations, a novel deep learning traffic forecasting framework based on graph attention network (GAT) and temporal convolutional network (TCN) is presented in this paper, termed as graph attention temporal convolutional networks (GATCN). More specifically, GATCN deal with the spatial features by GAT, and the temporal features by TCN. The layer fused by GAT and TCN enables the proposed model to learn the spatio-temporal characteristics that lie in traffic flow, while considering exogenous factors. In addition, nodes in the graph can capture the information of their neighborhoods by stacking multiple layers. Precision and robustness of the proposed method have been evaluated through testing on the real-world dataset. Results show that the proposed model outperforms other baselines.
With the success of deep learning in classifying short trimmed videos, more attention has been focused on temporally segmenting and classifying activities in long untrimmed videos. State-of-the-art ...approaches for action segmentation utilize several layers of temporal convolution and temporal pooling. Despite the capabilities of these approaches in capturing temporal dependencies, their predictions suffer from over-segmentation errors. In this paper, we propose a multi-stage architecture for the temporal action segmentation task that overcomes the limitations of the previous approaches. The first stage generates an initial prediction that is refined by the next ones. In each stage we stack several layers of dilated temporal convolutions covering a large receptive field with few parameters. While this architecture already performs well, lower layers still suffer from a small receptive field. To address this limitation, we propose a dual dilated layer that combines both large and small receptive fields. We further decouple the design of the first stage from the refining stages to address the different requirements of these stages. Extensive evaluation shows the effectiveness of the proposed model in capturing long-range dependencies and recognizing action segments. Our models achieve state-of-the-art results on three datasets: 50Salads, Georgia Tech Egocentric Activities (GTEA), and the Breakfast dataset.
Accurate and rapid load forecasting for industrial customers has been playing a crucial role in modern power systems. Due to the variability of industrial customers' activities, individual industrial ...loads are usually too volatile to forecast accurately. In this paper, a short-term load forecasting model for industrial customers based on the Temporal Convolutional Network (TCN) and Light Gradient Boosting Machine (LightGBM) is proposed. Firstly, a fixed-length sliding time window method is adopted to reconstruct the electrical features. Next, the TCN is utilized to extract the hidden information and long-term temporal relationships in the input features including electrical features, a meteorological feature and date features. Further, a state-of-the-art LightGBM capable of forecasting industrial customers' loads is adopted. The effectiveness of the proposed model is demonstrated by using datasets from different industries in China, Australia and Ireland. Multiple experiments and comparisons with existing models show that the proposed model provides accurate load forecasting results.
Vascular interventional surgery (VIS) robot is a surgical treatment plan that effectively protects surgeons from X-ray radiation. However, the master-slave control method cuts off the surgeons' ...natural force feedback, which increases the risk of surgical safety. Most VIS robotic systems use force sensors placed at the proximal end of guidewire to achieve force feedback, but due to the non-rigidity of the guidewire and the influence of mechanism friction, the proximal force collected has certain errors. In addition, the current VIS robotic systems are also insufficient in functionality, and cannot simultaneously complete the delivery of multiple surgical instruments. To solve the above mechanism design and force feedback challenges, a novel VIS robotic system equipped with force sensing mechanism is developed in this study. In addition, a temporal convolutional network (TCN) for the guidewire distal force prediction and an enhanced interactive force feedback strategy are proposed to improve the safety of the robotic system. Finally, combining the developed robotic system and the enhanced interactive force feedback strategy, a series of performance evaluations and model experiments are carried out. The results of the study demonstrate the effectiveness of the developed robotic system and the feasibility of the enhanced interactive force feedback strategy in improving surgical safety.
Although the accurate potential for growth prediction is very important for Government grants and contributions programs to better support Small and Medium-sized Enterprises (SMEs), it is a ...challenging task due to the data heterogeneity (both structured data and free text data bilingual in English and French), the class imbalance issue, and the difficulties in efficient feature learning. To address these challenges, this paper presents a novel BERT-TCN model for portfolio predictions in government funding programs, with the following key contributions. First, we describe the application of a novel architecture to a prediction task involving sequential, structured, partially quantitative input data and free text input data. Specifically, our novel model predicts the growth of firms receiving government funding for innovation. Our model also deals with class imbalance in the data and the difficulties in efficient feature learning. Our model integrates a Transformer model, i.e., BERT, for text modeling with a Temporal Convolutional Network (TCN) for sequential prediction. Second, we also developed various performance evaluation criteria in Section 4.3, allowing comprehensive assessments of the proposed approach from both the machine learning perspective and funding program-specific perspective. Third, the importance of features (both text and numerical features) is quantified and evaluated, allowing insights into how different features contribute to the prediction and explainability of the proposed model. The proposed approach is trained and tested on a large dataset from a rich database, demonstrating that the proposed approach can greatly help individual human experts improve their results.
•Proposing a novel BERT-TCN model for text modeling and temporal prediction.•Designing various evaluation criteria for comprehensive assessments of the model.•Exploring explainability of the model by ablation studies of various input features.
Multimodal sentiment analysis aims to extract and integrate information from different modalities to accurately identify the sentiment expressed in multimodal data. How to effectively capture the ...relevant information within a specific modality and how to fully exploit the complementary information among multiple modalities are two major challenges in multimodal sentiment analysis. Traditional approaches fail to obtain the global contextual information of long time-series data when extracting unimodal temporal features, and they usually fuse the features from multiple modalities with the same method and ignore the correlation between different modalities when modeling inter-modal interactions. In this paper, we first propose an Attentional Temporal Convolutional Network (ATCN) to extract unimodal temporal features for enhancing the feature representation ability, then introduce a Multi-layer Feature Fusion (MFF) model to improve the effectiveness of multimodal fusion, which fuses the different-level features by different methods according to the correlation coefficient between the features, and cross-modal multi-head attention is used to fully explore the potential relationship between the low-level features. The experimental results on SIMS and CMU-MOSI datasets show that the proposed model achieves superior performance on sentiment analysis tasks compared to state-of-the-art baselines.
Despite the success of data-driven converter fault diagnosis methods, interpretability and generalizability limit the further promotion of data-driven methods in industrial applications. Therefore, ...to improve the accuracy in face of out-of-distribution problems and increase confidence of power converter fault diagnosis, it is essential to understand the change and decision mechanism inside the deep model. First, we construct a general temporal convolutional network to visualize the diagnostic process, which has been proven effective in power converter fault diagnosis. Then, the effect of hyperparameters on generalizability is analyzed under typical power converter disturbances. Finally, the concern area of the model for the current in the fault decision is interpreted intuitively by gradient-weighted class activation mapping and the feature maps generated by the different channels are analyzed from multiple perspectives. The visualization results help to understand the complex structure of neural networks and can support the design of model to improve generalizability.
PDF
