Search results

To understand the meaning of a sentence, humans can focus on important words in the sentence, which reflects our eyes staying on each word in different gaze time or times. Thus, some studies utilize eye-tracking values to optimize the attention mechanism in deep learning models. But these studies lack to explain the rationality of this approach. Whether the attention mechanism possesses this feature of human reading needs to be explored.

The authors conducted experiments on a sentiment classification task. Firstly, they obtained eye-tracking values from two open-source eye-tracking corpora to describe the feature of human reading. Then, the machine attention values of each sentence were learned from a sentiment classification model. Finally, a comparison was conducted to analyze machine attention values and eye-tracking values.

Through experiments, the authors found the attention mechanism can focus on important words, such as adjectives, adverbs and sentiment words, which are valuable for judging the sentiment of sentences on the sentiment classification task. It possesses the feature of human reading, focusing on important words in sentences when reading. Due to the insufficient learning of the attention mechanism, some words are wrongly focused. The eye-tracking values can help the attention mechanism correct this error and improve the model performance.

Our research not only provides a reasonable explanation for the study of using eye-tracking values to optimize the attention mechanism but also provides new inspiration for the interpretability of attention mechanism.

This paper aims to propose a classification method for steel strip surface defects based on a mixed attention mechanism to achieve fast and accurate classification performance. The traditional method of classifying surface defects of hot-rolled steel strips has the problems of low recognition accuracy and low efficiency in the industrial complex production environment.

The authors selected min–max scaling comparison method to filter the training results of multiple network models on the steel strip surface defect data set. Then, the best comprehensive performance model EfficientNet-B0 was refined. Based on this, the authors proposed two mixed attention addition methods, which include squeeze-excitation spatial mixed module and multilayer mixed attention mechanism (MMAM) module, respectively.

With these two methods, the authors achieved 96.72% and 97.70% recognition accuracy on the steel strip data set after data augmentation for adapting to the complex production environment, respectively. Using the transfer learning method, the EfficientNet-B0 based on MMAM obtained 100% recognition accuracy.

This study not only focuses on improving the recognition accuracy of the network model itself but also considers other performance indicators of the network, which are rarely considered by many researchers. The authors further improve the intelligent production technique and address this issue. Both methods proposed in this paper can be applied to embedded equipment, which can effectively improve steel strip factory production efficiency and reduce material and time loss.

The current studies on remaining useful life (RUL) prediction mainly rely on convolutional neural networks (CNNs) and long short-term memories (LSTMs) and do not take full advantage of the attention mechanism, resulting in lack of prediction accuracy. To further improve the performance of the above models, this study aims to propose a novel end-to-end RUL prediction framework, called convolutional recurrent attention network (CRAN) to achieve high accuracy.

The proposed CRAN is a CNN-LSTM-based model that effectively combines the powerful feature extraction ability of CNN and sequential processing capability of LSTM. The channel attention mechanism, spatial attention mechanism and LSTM attention mechanism are incorporated in CRAN, assigning different attention coefficients to CNN and LSTM. First, features of the bearing vibration data are extracted from both time and frequency domain. Next, the training and testing set are constructed. Then, the CRAN is trained offline using the training set. Finally, online RUL estimation is performed by applying data from the testing set to the trained CRAN.

CNN-LSTM-based models have higher RUL prediction accuracy than CNN-based and LSTM-based models. Using a combination of max pooling and average pooling can reduce the loss of feature information, and in addition, the structure of the serial attention mechanism is superior to the parallel attention structure. Comparing the proposed CRAN with six different state-of-the-art methods, for the predicted results of two testing bearings, the proposed CRAN has an average reduction in the root mean square error of 57.07/80.25%, an average reduction in the mean absolute error of 62.27/85.87% and an average improvement in score of 12.65/6.57%.

This article provides a novel end-to-end rolling bearing RUL prediction framework, which can provide a reference for the formulation of bearing maintenance programs in the industry.

In a classroom, a support vector machines model with a linear kernel, a neural network and the k-nearest neighbors algorithm failed to detect simulated money laundering accounts generated from the Panama papers data set of the offshore leak database. This study aims to resolve this failure.

Build a graph attention network having three modules as a new money laundering detection tool. A feature extraction module encodes these input data to create a weighted graph structure. In it, directed edges and their end vertices denote financial transactions. Each directed edge has weights for storing the frequency of money transactions and other significant features. Social network metrics are features of nodes for characterizing an account’s roles in a money laundering typology. A graph attention module implements a self-attention mechanism for highlighting target nodes. A classification module further filters out such targets using the biased rectified linear unit function.

Resulted from the highlighting of nodes using a self-attention mechanism, the proposed graph attention network outperforms a Naïve Bayes classifier, the random forest method and a support vector machines model with a radial kernel in detecting money laundering accounts. The Naïve Bayes classifier produces second accurate classifications.

This paper develops a new money laundering detection tool, which outperforms existing methods. This new tool produces more accurate detections of money laundering, perfects warns of money laundering accounts or links and provides sharp efficiency in processing financial transaction records without being afraid of their amount.

Despite its significance, research on how attribute framing affects ordering decisions in dual sourcing remains insufficient. Hence, this study investigated the effects of attribute framing in a sourcing task involving certain and uncertain qualities of two suppliers and analysed the role of attention with respect to suppliers' information in framing effects.

The impacts of attribute framing on sourcing decisions were demonstrated in two online between-subject (2 × 2 factorial) experimental studies involving professional samples. Study 2 was an eye-tracking experiment.

In Study 1 (N = 251), participants presented with a “high-quality” rather than a “low-quality” frame made different sourcing decisions, opting for larger percentage of order(s) from a supplier under the “high-quality” frame. This pattern holds true for suppliers who differ in risk. This finding was replicated in Study 2 (N = 129). Attention asymmetry related to the information on supplier quality contributes to this effect. Attention directed towards information regarding the supplier's quality under a positive frame mediated the relationship between attribute framing and sourcing decisions.

Highlighting the positive attributes of a risky supplier is essential when ordering from the risky supplier is an optimal decision. It is advantageous for suppliers to highlight positive rather than negative attributes when describing the quality of their components against others.

This is the first study to examine the effect of attention on the relationship between attribute framing and dual sourcing. This presents a new behavioural perspective wherein managers' attention to information plays a vital role.

This paper aims to address the nature of docility in organizations, its practical role in attention scarcity and knowledge diffusion in complex organizations and the management implications for organizational learning and innovation to improve knowledge management.

This paper examines knowledge organizations from the perspective of human resource strategies, their role in information abundance and attention scarcity and techniques to enhance docility mechanisms at different levels of the organization to increase innovation and performance.

This paper, in reviewing the organization literature on attention scarcity, addresses the shortage of studies linking the need for docility – the desire to learn from workers and the desire to teach – in personnel practices of knowledge firms, where intense social interaction, social feedback and social learning are the norms.

Knowledge management – scanning, creation, coordination, interpreting, transfer and integration – may well be the basis of competitive advantage, based on human resource strategies to mobilize explicit and tacit knowledge via docility mechanisms, including mentoring, teamwork, coaching and deep collaboration.

Decades ago, Herbert A. Simon introduced this new concept, docility, which is now central to knowledge organizations that face information abundance and attention scarcity. Knowledge organizations require tools of docility to align human resource strategies to both strategic management and operational functions to enhance teaching and learning in design structures that are time-constrained.

In novelty detection, the autoencoder based image reconstruction strategy is one of the mainstream solutions. The basic idea is that once the autoencoder is trained on normal data, it has a low reconstruction error on normal data. However, when faced with complex natural images, the conventional pixel-level reconstruction becomes poor and does not show the promising results. This paper aims to provide a new method for improving the performance of novelty detection based autoencoder.

To solve the problem that conventional pixel-level reconstruction cannot effectively extract the global semantic information of the image, a novel model with the combination of attention mechanism and self-supervised learning method is proposed. First, an auxiliary task, reconstruct rotated image, is set to enable the network to learn global semantic feature information. Then, the channel attention mechanism is introduced to perform adaptive feature refinement on the intermediate feature map to optimize the correspondingly passed feature map.

Experimental results on three public data sets show that the proposed method has potential performance for novelty detection.

This study explores the ability of self-supervised learning methods and attention mechanism to extract features on a single class of images. In this way, the performance of novelty detection can be improved.

Attention is one of the most important factors to affect the academic performance of students. Effectively analyzing students' attention in class can promote teachers' precise teaching and students' personalized learning. To intelligently analyze the students' attention in classroom from the first-person perspective, this paper proposes a fusion model based on gaze tracking and object detection. In particular, the proposed attention analysis model does not depend on any smart equipment.

Given a first-person view video of students' learning, the authors first estimate the gazing point by using the deep space–time neural network. Second, single shot multi-box detector and fast segmentation convolutional neural network are comparatively adopted to accurately detect the objects in the video. Third, they predict the gazing objects by combining the results of gazing point estimation and object detection. Finally, the personalized attention of students is analyzed based on the predicted gazing objects and the measurable eye movement criteria.

A large number of experiments are carried out on a public database and a new dataset that is built in a real classroom. The experimental results show that the proposed model not only can accurately track the students' gazing trajectory and effectively analyze the fluctuation of attention of the individual student and all students but also provide a valuable reference to evaluate the process of learning of students.

The contributions of this paper can be summarized as follows. The analysis of students' attention plays an important role in improving teaching quality and student achievement. However, there is little research on how to automatically and intelligently analyze students' attention. To alleviate this problem, this paper focuses on analyzing students' attention by gaze tracking and object detection in classroom teaching, which is significant for practical application in the field of education. The authors proposed an effectively intelligent fusion model based on the deep neural network, which mainly includes the gazing point module and the object detection module, to analyze students' attention in classroom teaching instead of relying on any smart wearable device. They introduce the attention mechanism into the gazing point module to improve the performance of gazing point detection and perform some comparison experiments on the public dataset to prove that the gazing point module can achieve better performance. They associate the eye movement criteria with visual gaze to get quantifiable objective data for students' attention analysis, which can provide a valuable basis to evaluate the learning process of students, provide useful learning information of students for both parents and teachers and support the development of individualized teaching. They built a new database that contains the first-person view videos of 11 subjects in a real classroom and employ it to evaluate the effectiveness and feasibility of the proposed model.

Automatic segmentation of brain tumor from medical images is a challenging task because of tumor's uneven and irregular shapes. In this paper, the authors propose an attention-based nested segmentation network, named DAU-Net. In total, two types of attention mechanisms are introduced to make the U-Net network focus on the key feature regions. The proposed network has a deep supervised encoder–decoder architecture and a redesigned dense skip connection. DAU-Net introduces an attention mechanism between convolutional blocks so that the features extracted at different levels can be merged with a task-related selection.

In the coding layer, the authors designed a channel attention module. It marks the importance of each feature graph in the segmentation task. In the decoding layer, the authors designed a spatial attention module. It marks the importance of different regional features. And by fusing features at different scales in the same coding layer, the network can fully extract the detailed information of the original image and learn more tumor boundary information.

To verify the effectiveness of the DAU-Net, experiments were carried out on the BRATS 2018 brain tumor magnetic resonance imaging (MRI) database. The segmentation results show that the proposed method has a high accuracy, with a Dice similarity coefficient (DSC) of 89% in the complete tumor, which is an improvement of 8.04 and 4.02%, compared with fully convolutional network (FCN) and U-Net, respectively.

The experimental results show that the proposed method has good performance in the segmentation of brain tumors. The proposed method has potential clinical applicability.

Perception has been identified as the main cause underlying most autonomous vehicle related accidents. As the key technology in perception, deep learning (DL) based computer vision models are generally considered to be black boxes due to poor interpretability. These have exacerbated user distrust and further forestalled their widespread deployment in practical usage. This paper aims to develop explainable DL models for autonomous driving by jointly predicting potential driving actions with corresponding explanations. The explainable DL models can not only boost user trust in autonomy but also serve as a diagnostic approach to identify any model deficiencies or limitations during the system development phase.

This paper proposes an explainable end-to-end autonomous driving system based on “Transformer,” a state-of-the-art self-attention (SA) based model. The model maps visual features from images collected by onboard cameras to guide potential driving actions with corresponding explanations, and aims to achieve soft attention over the image’s global features.

The results demonstrate the efficacy of the proposed model as it exhibits superior performance (in terms of correct prediction of actions and explanations) compared to the benchmark model by a significant margin with much lower computational cost on a public data set (BDD-OIA). From the ablation studies, the proposed SA module also outperforms other attention mechanisms in feature fusion and can generate meaningful representations for downstream prediction.

In the contexts of situational awareness and driver assistance, the proposed model can perform as a driving alarm system for both human-driven vehicles and autonomous vehicles because it is capable of quickly understanding/characterizing the environment and identifying any infeasible driving actions. In addition, the extra explanation head of the proposed model provides an extra channel for sanity checks to guarantee that the model learns the ideal causal relationships. This provision is critical in the development of autonomous systems.