Search

Accurate delineation of key waveforms in an ECG is a critical step in extracting relevant features to support the diagnosis and treatment of heart conditions. Although deep learning based methods using segmentation models to locate P, QRS, and T waves have shown promising results, their ability to handle arrhythmias has not been studied in any detail. In this paper we investigate the effect of arrhythmias on delineation quality and develop strategies to improve performance in such cases. We introduce a U-Net-like segmentation model for ECG delineation with a particular focus on diverse arrhythmias. This is followed by a post-processing algorithm which removes noise and automatically determines the boundaries of P, QRS, and T waves. Our model has been trained on a diverse dataset and evaluated against the LUDB and QTDB datasets to show strong performance, with F1-scores exceeding 99% for QRS and T waves, and over 97% for P waves in the LUDB dataset. Furthermore, we assess various models across a wide array of arrhythmias and observe that models with a strong performance on standard benchmarks may still perform poorly on arrhythmias that are underrepresented in these benchmarks, such as tachycardias. We propose solutions to address this discrepancy. [ABSTRACT FROM AUTHOR]

We provide new algorithms for two tasks relating to heterogeneous tabular datasets: clustering, and synthetic data generation. Tabular datasets typically consist of heterogeneous data types (numerical, ordinal, categorical) in columns, but may also have hidden cluster structure in their rows: for example, they may be drawn from heterogeneous (geographical, socioeconomic, methodological) sources, such that the outcome variable they describe (such as the presence of a disease) may depend not only on the other variables but on the cluster context. Moreover, sharing of biomedical data is often hindered by patient confidentiality laws, and there is current interest in algorithms to generate synthetic tabular data from real data, for example via deep learning. We demonstrate a novel EM-based clustering algorithm, MMM ("Madras Mixture Model"), that outperforms standard algorithms in determining clusters in synthetic heterogeneous data, and recovers structure in real data. Based on this, we demonstrate a synthetic tabular data generation algorithm, MMMsynth, that pre-clusters the input data, and generates cluster-wise synthetic data assuming cluster-specific data distributions for the input columns. We benchmark this algorithm by testing the performance of standard ML algorithms when they are trained on synthetic data and tested on real published datasets. Our synthetic data generation algorithm outperforms other literature tabular-data generators, and approaches the performance of training purely with real data. [ABSTRACT FROM AUTHOR]

Brain segmentation from neonatal MRI images is a very challenging task due to large changes in the shape of cerebral structures and variations in signal intensities reflecting the gestational process. In this context, there is a clear need for segmentation techniques that are robust to variations in image contrast and to the spatial configuration of anatomical structures. In this work, we evaluate the potential of synthetic learning, a contrast‐independent model trained using synthetic images generated from the ground truth labels of very few subjects. We base our experiments on the dataset released by the developmental Human Connectome Project, for which high‐quality images are available for more than 700 babies aged between 26 and 45 weeks postconception. First, we confirm the impressive performance of a standard UNet trained on a few volumes, but also confirm that such models learn intensity‐related features specific to the training domain. We then confirm the robustness of the synthetic learning approach to variations in image contrast. However, we observe a clear influence of the age of the baby on the predictions. We improve the performance of this model by enriching the synthetic training set with realistic motion artifacts and over‐segmentation of the white matter. Based on extensive visual assessment, we argue that the better performance of the model trained on real T2w data may be due to systematic errors in the ground truth. We propose an original experiment allowing us to show that learning from real data will reproduce any systematic bias affecting the training set, while synthetic models can avoid this limitation. Overall, our experiments confirm that synthetic learning is an effective solution for segmenting neonatal brain MRI. Our adapted synthetic learning approach combines key features that will be instrumental for large multisite studies and clinical applications. [ABSTRACT FROM AUTHOR]

Recent developments in the field of non-local attention (NLA) have led to a renewed interest in self-similarity-based single image super-resolution (SISR). Researchers usually use the NLA to explore non-local self-similarity (NSS) in SISR and achieve satisfactory reconstruction results. However, a surprising phenomenon that the reconstruction performance of the standard NLA is similar to that of the NLA with randomly selected regions prompted us to revisit NLA. In this paper, we first analyzed the attention map of the standard NLA from different perspectives and discovered that the resulting probability distribution always has full support for every local feature, which implies a statistical waste of assigning values to irrelevant non-local features, especially for SISR which needs to model long-range dependence with a large number of redundant non-local features. Based on these findings, we introduced a concise yet effective soft thresholding operation to obtain high-similarity-pass attention (HSPA), which is beneficial for generating a more compact and interpretable distribution. Furthermore, we derived some key properties of the soft thresholding operation that enable training our HSPA in an end-to-end manner. The HSPA can be integrated into existing deep SISR models as an efficient general building block. In addition, to demonstrate the effectiveness of the HSPA, we constructed a deep high-similarity-pass attention network (HSPAN) by integrating a few HSPAs in a simple backbone. Extensive experimental results demonstrate that HSPAN outperforms state-of-the-art approaches on both quantitative and qualitative evaluations. Our code and a pre-trained model were uploaded to GitHub (https://github.com/laoyangui/HSPAN) for validation. [ABSTRACT FROM AUTHOR]

Increasing spectrum efficiency in new-generation communication networks is possible by raising the operating frequencies or increasing the number of serving cells. Both of the solutions indirectly lead to a decrease in the cell size. By shrinking the coverage area of the cells, the mobility management becomes a critical issue in order to provide seamless connectivity. The main tool for mobility management in connected mode is the handover procedure, which conventionally is triggered based on the user equipment's (UE's) measurements of serving and neighboring cells' signal quality. The performance of the methods provided in the 3GPP standard for setting up the handover is based on a comparison between the quality of the received signal from the serving cell and the neighbor cells, which is severely sensitive to the appropriate setting of threshold values and waiting time for setting up the handover. The remarkable methods in the literature to improve the handover performance in the 3GPP standard either utilize positional information or are based on complex algorithms. Positional information parameters such as speed, location, direction of movement, etc., need extra measurement modules. Also, complex algorithms like deep learning are not suitable for a wide range of active devices in 5 G networks with limited resources. In this paper, a novel fuzzy logic-based method is proposed to trigger the handover procedure based on estimated serving cell's and neighbor cells' radio link quality (RLQ) values. The proposed system consists of two stages. A second-order regressor beside a simple fuzzy logic system is introduced to predict the serving cell's and neighbor cells' RLQ. The final handover trigger decision is made with the help of another cascade fuzzy logic system, which is responsible for eliminating too-early, too-late, and ping-pong handovers. Considering the uncertainty handling feature of the interval type II fuzzy logic systems versus type I, we implement both methods and compare their results. Finally, we simulate the proposed algorithm using the ns-3 LTE module with a very tight setting for detecting radio link failure to meet 5 G strict standards. Proposed method succeeded in improving the performance of the handover process at high-speed scenarios by 50%, only with the help of radio link quality information. The main advantage of the proposed method is the proper management of the handover procedure, independent of UE's velocity, as well as its simple structure with few rules, which makes it suitable for use in UAV and IoT devices. [ABSTRACT FROM AUTHOR]

Objectives: Development of automated segmentation models enabling standardized volumetric quantification of fibroglandular tissue (FGT) from native volumes and background parenchymal enhancement (BPE) from subtraction volumes of dynamic contrast-enhanced breast MRI. Subsequent assessment of the developed models in the context of FGT and BPE Breast Imaging Reporting and Data System (BI-RADS)-compliant classification. Methods: For the training and validation of attention U-Net models, data coming from a single 3.0-T scanner was used. For testing, additional data from 1.5-T scanner and data acquired in a different institution with a 3.0-T scanner was utilized. The developed models were used to quantify the amount of FGT and BPE in 80 DCE-MRI examinations, and a correlation between these volumetric measures and the classes assigned by radiologists was performed. Results: To assess the model performance using application-relevant metrics, the correlation between the volumes of breast, FGT, and BPE calculated from ground truth masks and predicted masks was checked. Pearson correlation coefficients ranging from 0.963 ± 0.004 to 0.999 ± 0.001 were achieved. The Spearman correlation coefficient for the quantitative and qualitative assessment, i.e., classification by radiologist, of FGT amounted to 0.70 (p < 0.0001), whereas BPE amounted to 0.37 (p = 0.0006). Conclusions: Generalizable algorithms for FGT and BPE segmentation were developed and tested. Our results suggest that when assessing FGT, it is sufficient to use volumetric measures alone. However, for the evaluation of BPE, additional models considering voxels' intensity distribution and morphology are required. Critical relevance statement: A standardized assessment of FGT density can rely on volumetric measures, whereas in the case of BPE, the volumetric measures constitute, along with voxels' intensity distribution and morphology, an important factor. Key points: • Our work contributes to the standardization of FGT and BPE assessment. • Attention U-Net can reliably segment intricately shaped FGT and BPE structures. • The developed models were robust to domain shift. [ABSTRACT FROM AUTHOR]

The ability to sequentially learn multiple tasks without forgetting is a key skill of biological brains, whereas it represents a major challenge to the field of deep learning. To avoid catastrophic forgetting, various continual learning (CL) approaches have been devised. However, these usually require discrete task boundaries. This requirement seems biologically implausible and often limits the application of CL methods in the real world where tasks are not always well defined. Here, we take inspiration from neuroscience, where sparse, non-overlapping neuronal representations have been suggested to prevent catastrophic forgetting. As in the brain, we argue that these sparse representations should be chosen on the basis of feed forward (stimulus-specific) as well as top-down (context-specific) information. To implement such selective sparsity, we use a bio-plausible form of hierarchical credit assignment known as Deep Feedback Control (DFC) and combine it with a winner-take-all sparsity mechanism. In addition to sparsity, we introduce lateral recurrent connections within each layer to further protect previously learned representations. We evaluate the new sparse-recurrent version of DFC on the split-MNIST computer vision benchmark and show that only the combination of sparsity and intra-layer recurrent connections improves CL performance with respect to standard backpropagation. Our method achieves similar performance to well-known CL methods, such as Elastic Weight Consolidation and Synaptic Intelligence, without requiring information about task boundaries. Overall, we showcase the idea of adopting computational principles from the brain to derive new, task-free learning algorithms for CL. [ABSTRACT FROM AUTHOR]

Diabetic retinopathy, which is extreme visual blindness due to diabetes, has become an alarming issue worldwide. Early and accurate detection of DR is necessary to prevent the progression and reduce the risk of blindness. Recently, many approaches for DR detection have been proposed in the literature. Among them, deep neural networks (DNNs), especially Convolutional Neural Network (CNN) models, have become the most offered approach. However, designing and training new CNN architectures from scratch is a troublesome and labor-intensive task, particularly for medical images. Moreover, it requires training tremendous amounts of parameters. Therefore, transfer learning approaches as pre-trained models have become more prevalent in the last few years. Accordingly, in this study, 43 publications based on DNN and Transfer Learning approaches for DR detection between 2016 and 2021 are reviewed. The reviewed papers are summarized in 4 figures and 10 tables that present detailed information about 29 pre-trained CNN models, 13 DR data sets, and standard performance metrics. [ABSTRACT FROM AUTHOR]

Parkinson's disease is one of the most common degenerative conditions that affect people aged 60 and older. The illness is normally diagnosed by clinical indicators that develop as a variety of movement symptoms and medical observations. Conventional diagnostic methods, on the other hand, rely on the detection of small motions, which are notoriously difficult to pin down with absolute precision. This makes it possible for these methods to lend themselves to subjective interpretations. This is because traditional diagnostic methods rely on the interpretation of motions. Image categorization is performed using an altered version of the ResNeXt model in the proposed model. The proposed ResNeXt extends version of ResNet architecture by introducing a new block called the "cardinality block." The cardinality block consists of multiple parallel branches, each with its own set of convolutional layers. These branches are then combined by concatenation before being passed to the next layer. The key idea behind the cardinality block is to increase the capacity of the network without significantly increasing the number of parameters. By using parallel branches with different filter sizes, ResNeXt is able to capture a wider range of features in the input image, leading to better performance on image classification tasks. In order to enhance the performance of the standard ResNeXt model, certain extra dense and dropout layers have been included. The size of the final model is reduced by pruning the model in order to improve the efficiency of the network connections and minimize the overall size of the model. The proposed method is compared to a number of deep learning models that are already in use, and it is shown that the acquired results are superior to those of the deep learning models that are already in use. [ABSTRACT FROM AUTHOR]

Intent classification, to identify the speaker's intention, and slot filling, to label each token with a semantic type, are critical tasks in natural language understanding. Traditionally the two tasks have been addressed independently. More recently joint models that address the two tasks together have achieved state-of-the-art performance for each task and have shown there exists a strong relationship between the two. In this survey, we bring the coverage of methods up to 2021 including the many applications of deep learning in the field. As well as a technological survey, we look at issues addressed in the joint task and the approaches designed to address these issues. We cover datasets, evaluation metrics, and experiment design and supply a summary of reported performance on the standard datasets. [ABSTRACT FROM AUTHOR]

Deep reinforcement learning (DRL) has been proven as a powerful approach for robot navigation over the past few years. DRL-based navigation does not require the pre-construction of a map, instead, high-performance navigation skills can be learned from trial-and-error experiences. However, recent DRL-based approaches mostly focus on a fixed navigation target. It is noted that when navigating to a moving target without maps, the performance of the standard RL structure drops dramatically on both the success rate and path efficiency. To address the mapless navigation problem with moving target, the predictive hierarchical DRL (pH-DRL) framework is proposed by integrating the long-term trajectory prediction to provide a cost-effective solution. In the proposed framework, the lower-level policy of the RL agent learns robot control actions to a specified goal, and the higher-level policy learns to make long-range planning of shorter navigation routes by sufficiently exploiting the predicted trajectories. By means of making decisions over two level of policies, the pH-DRL framework is robust to the unavoidable errors in long-term predictions. With the application of deep deterministic policy gradient (DDPG) for policy optimization, the pH-DDPG algorithm is developed based on the pH-DRL structure. Finally, through comparative experiments on the Gazebo simulator with several variants of the DDPG algorithm, the results demonstrate that the pH-DDPG outperforms other algorithms and achieves a high success rate and efficiency even though the target moves fast and randomly. [ABSTRACT FROM AUTHOR]

Object detection from underwater sea images based on deep learning techniques provides preferable results in a controlled environment. Yet, these techniques experience some challenges in detecting underwater objects due to color distortion, noise, and scattering. Hence, enhancing the underwater imaginary is important for accurately determining the objects under water. This research presents a deep learning approach for perceiving underwater objects from enhanced underwater images. Very Deep Super-Resolution Network (VDSR), which exhibits a higher visual quality, is utilized for improving the underwater image quality, thereby it is sufficient for object detection. Then, the object is detected from the enhanced underwater image through the proposed Border Collie Flamingo optimization-based deep CNN classifier (BCFO-based deep CNN). The developed BCFO-based algorithm is the main highlight of the research, which effectively tunes the classifier's hyperparameter. The evaluation is established using the UIEB and DUO datasets on the basis of performance standards, such as specificity, accuracy, and sensitivity. When the training percentage is 80 and the K -fold is 10, the suggested model achieved accuracy results of 93.89% and 95.24%, sensitivity results of 95.93 and 97.29%, and specificity results of 98.64% and 99%, which is very efficient compared to some existing approaches. [ABSTRACT FROM AUTHOR]

Rising stars are the researchers who are relatively new to the research area and have published fewer research articles, but their research work is of such standard that they have the potential to be top researchers in near future. Research work on the evaluation of researchers and prediction of rising stars is getting attention because it can be useful for selecting capable candidates for the jobs, hiring young faculty members for institutes, and seeking reviewers for journals and conferences and members for different committees. In this research study, the authors address the research problem of finding rising stars and propose novel features in diverse feature sets of three categories: article, author, and venue. The real-world data set has been extracted, preprocessed, and used from the Web of Science for empirical analysis. Several diverse supervised machine learning, ensemble learning algorithms, and deep learning are applied to the data set. The results, using classifiers, are compared based on standard performance evaluation measures to reveal the significance of the proposed as well as existing features. It also shows that the novel features play a significant role in finding rising stars. The ensemble-based machine learning classifier generalized linear model outperforms all other classifiers and gives the highest accuracy and F-measure compared to other models and the existing studies in the relevant literature. [ABSTRACT FROM AUTHOR]

Motivation Sequence-based deep learning approaches have been shown to predict a multitude of functional genomic readouts, including regions of open chromatin and RNA expression of genes. However, a major limitation of current methods is that model interpretation relies on computationally demanding post hoc analyses, and even then, one can often not explain the internal mechanics of highly parameterized models. Here, we introduce a deep learning architecture called totally interpretable sequence-to-function model (tiSFM). tiSFM improves upon the performance of standard multilayer convolutional models while using fewer parameters. Additionally, while tiSFM is itself technically a multilayer neural network, internal model parameters are intrinsically interpretable in terms of relevant sequence motifs. Results We analyze published open chromatin measurements across hematopoietic lineage cell-types and demonstrate that tiSFM outperforms a state-of-the-art convolutional neural network model custom-tailored to this dataset. We also show that it correctly identifies context-specific activities of transcription factors with known roles in hematopoietic differentiation, including Pax5 and Ebf1 for B-cells, and Rorc for innate lymphoid cells. tiSFM's model parameters have biologically meaningful interpretations, and we show the utility of our approach on a complex task of predicting the change in epigenetic state as a function of developmental transition. Availability and implementation The source code, including scripts for the analysis of key findings, can be found at https://github.com/boooooogey/ATAConv , implemented in Python. [ABSTRACT FROM AUTHOR]

Deep convolutional networks are data hungry learners and, to compensate for the limited amount of available data, various augmentation methods have been proposed. While the initial approaches aimed to fill the space around existing data points, more recent methods use the "mixing" procedure, which is a convex combination between points, being both holistic and local (from a spatial point of view). Although the mixing techniques improved the performance for standard benchmarks, we do notice that they are less effective for problems that exhibit poor class separation. For these scenarios, we propose a soft labeling technique based on distances to class prototypes, as extracted on an intermediate CNN layer. The mixing is performed between relative positions in the clustering space. The method, named SoftClusterMix, is shown to be competitive on standard image classification benchmarks and leads to significantly improved accuracy for problems where there is a poor class separation. We report better performance in three such categories: face expression recognition, aesthetic image classification and painting style classification. Ablation tests allow deep insights of the method. [ABSTRACT FROM AUTHOR]

Mobile edge computing (MEC) is an emerging paradigm that decreases the computational burden of mobiles by task offloading. MEC is regarded as an effective method to offer computing capacities in close proximities to mobile users. The major issue in MEC is how to offload the heterogeneous task of mobile apps effectively from the user equipment to the MEC host. Some of the existing techniques contain feeble adaptability to new circumstance due to minimum sample effectiveness and require complete retraining. The purpose of MEC is to efficiently solve offloading issues such as network load and latency. In this paper, a deep belief network is proposed for solving the offloading issue in the clusters of numerous service node and several dependencies for a mobile task in huge-scale heterogeneous mobile edge computing. The deep belief network's weight is tuned optimally using Adaptive Harris Hawks optimization (AHHO) algorithm for solving the offloading issue optimally in the MEC environment. An AHHO algorithm is employed to improve the search performances of the standard algorithm. The limitations of the standard algorithm are poor stability among exploitation as well as exploration. Hence, two schemes such as the Gaussian mutation scheme and the cuckoo search are combined so as to form an adaptive Harris hawks optimization to enhance this stability. The task offloading issues are implemented using Google cluster trace and iFogSim. Furthermore, the simulation results depict that the offloading scheme based on the deep belief network-based adaptive Harris Hawks optimization approach has better results with respect to measures such as load balancing, energy consumption, average execution time, and latency than any other approach. [ABSTRACT FROM AUTHOR]

Foreign species can deteriorate the environment and the economy of a country. To automatically monitor biosecurity threats at country borders, this paper investigates compact deep networks for accurate and real-time object segmentation for hyperspectral images. To this end, knowledge distillation (KD) approaches compress the model by distilling the knowledge of a large teacher network to a compact student network. However, when the student is over-compressed, the performance of standard KD methods degrades significantly due to the large capacity gap between the teacher and the student. This gap can be addressed by adding medium-sized teacher assistants, but training them incurs significant computation and hence is impractical. To address this problem, this paper proposes a new framework called Knowledge Distillation from Multi-head Teacher (KDM), which distills the knowledge of a multi-head teacher to the student. By encapsulating multiple teachers in a single network, our proposed KDM assists the learning of a very compact student and significantly reduces the training time. We also introduce Bio-HSI, a new large benchmark hyperspectral image dataset of 3,125 high-resolution images with dense segmentation ground truth. This new, large dataset can be expected to advance research on deep models for hyperspectral image segmentation. Evaluated on this dataset, the student trained via our KDM has 762 times fewer parameters than the state-of-the-art segmentation model (i.e., HRNet), while achieving competitive accuracy. [ABSTRACT FROM AUTHOR]

Existing person re-identification (Re-ID) methods usually rely heavily on large-scale thoroughly annotated training data. However, label noise is unavoidable due to inaccurate person detection results or annotation errors in real scenes. It is extremely challenging to learn a robust Re-ID model with label noise since each identity has very limited annotated training samples. To avoid fitting to the noisy labels, we propose to learn a prefatory model using a large learning rate at the early stage with a self-label refining strategy, in which the labels and network are jointly optimized. To further enhance the robustness, we introduce an online co-refining (CORE) framework with dynamic mutual learning, where networks and label predictions are online optimized collaboratively by distilling the knowledge from other peer networks. Moreover, it also reduces the negative impact of noisy labels using a favorable selective consistency strategy. CORE has two primary advantages: it is robust to different noise types and unknown noise ratios; it can be easily trained without much additional effort on the architecture design. Extensive experiments on Re-ID and image classification demonstrate that CORE outperforms its counterparts by a large margin under both practical and simulated noise settings. Notably, it also improves the state-of-the-art unsupervised Re-ID performance under standard settings. Code is available at https://github.com/mangye16/ReID-Label-Noise. [ABSTRACT FROM AUTHOR]

We present a new convolution layer for deep learning architectures which we call QuadConv — an approximation to continuous convolution via quadrature. Our operator is developed explicitly for use on non-uniform, mesh-based data, and accomplishes this by learning a continuous kernel that can be sampled at arbitrary locations. Moreover, the construction of our operator admits an efficient implementation which we detail and construct. As an experimental validation of our operator, we consider the task of compressing partial differential equation (PDE) simulation data from fixed meshes. We show that QuadConv can match the performance of standard discrete convolutions on uniform grid data by comparing a QuadConv autoencoder (QCAE) to a standard convolutional autoencoder (CAE). Further, we show that the QCAE can maintain this accuracy even on non-uniform data. In both cases, QuadConv also outperforms alternative unstructured convolution methods such as graph convolution. • A new convolution layer for deep learning architectures with unstructured data. • Convolutional operator learns a continuous kernel and a quadrature rule. • Convolutional layer used within autoencoders for unstructured PDE data compression. • Method leads to similar accuracy to CNN autoencoder on equivalent structured data. [ABSTRACT FROM AUTHOR]

Detection of lane region under the road boundary is an imperative module for intelligent vehicle system. Lane markings provide separate regions on the road for the vehicles to avoid the possibility of accidents. Existing methods in lane detection have limited performance using various sensor-based approaches such as Radar and LiDAR and have high operational costs. To achieve a steady and optimal lane detection, the vision-based lane region detection scheme VLDNet is proposed which utilizes a encoder-decoder network using semantic segmentation architecture. In this direction, a hybrid model using UNet and ResNet has been adopted, where UNet is used as a segmentation model and ResNet-50 is used for down-sampling the image and identifying the required features. These identified features have been then applied into UNet for up-sampling and decoding the segments of the images. The publicly available KITTI dataset have been accessed for experiments and validation of the proposed network. The method outperforms the existing state-of-the-art methods in lane region detection. The network achieves better performance using standard evaluation measures such as accuracy of 98.87%, the precision of 98.24%, recall of 96.55%, frequency weighted IoU of 97.78%, and MaxF score of 97.77%. [ABSTRACT FROM AUTHOR]

• New model for scene text VQA that jointly reasons about textual and visual modalities. • Attending on multi-modal features is better than attending separately to each modality. • Grid features from an object detection backbone proves to work well in this task. • The proposed model is up to 5 times faster than current state of the art models. • Combined with a standard VQA model reaches state of the art on mixed datasets. This paper presents a new model for the task of scene text visual question answering. In this task questions about a given image can only be answered by reading and understanding scene text. Current state of the art models for this task make use of a dual attention mechanism in which one attention module attends to visual features while the other attends to textual features. A possible issue with this is that it makes difficult for the model to reason jointly about both modalities. To fix this problem we propose a new model that is based on an single attention mechanism that attends to multi-modal features conditioned to the question. The output weights of this attention module over a grid of multi-modal spatial features are interpreted as the probability that a certain spatial location of the image contains the answer text to the given question. Our experiments demonstrate competitive performance in two standard datasets with a model that is × 5 faster than previous methods at inference time. Furthermore, we also provide a novel analysis of the ST-VQA dataset based on a human performance study. Supplementary material, code, and data is made available through this link. [ABSTRACT FROM AUTHOR]

Entity relation extraction is a fundamental task to detect a list of triplets, including two entities and the semantic relations between them. An overlapping relationship has caused the low performance of standard entity relation extraction in fishery in recent years. In this study, a novel entity relation extraction was proposed in fishery using the dual attention mechanism. First, a sentence classification and labeling strategy were selected to solve the difficulty in labeling overlapping relations in fishery standard texts. Second, an extraction model of standard entity relationship was established using a combination of dual attention, and Bidirectional Encoder Representations from Transformers-Bi-directional Long Short-Term Memory-Conditional Random Field (BERT-BiLSTM-CRF). Five components were included from the bottom to top: BERT layer, BiLSTM layer, word attention mechanism, sentence attention, and CRF output layer. In the pre-trained BERT model, the dual-layer two-way conversion and decoding were used to automatically learn sentence features for the vector representation. In the BiLSTM model, the context feature of the target entity was learned from the BERT output. The word and sentence level attention was utilized to optimize the weight of target words and sentences in the paragraph, including the noise removal, and the accuracy of relationship extraction. The CRF decoder was used to represent the output of the attention layer in the form of sequence tags. Finally, a comparative experiment was designed to verify the effectiveness of the model. The results show that better overall performance was achieved in the entity relationship extraction using the dual attention on the fishery standard interaction (DLOU-FSI) dataset, where the accuracy, recall, and F1 value were 92.67%, 92.31%, and 92.49%, respectively. The overlapping relations were effectively solved in the extraction of fishery standard relations. In addition, the recognition accuracy rates in the seven relation categories were all higher than others. The accuracy rate, recall rate, and F1 value were greatly improved by more than 90% in the quotation relationship, regulation, release, proposal, drafting, and centralization. Nevertheless, the recall rate of the improved model was dropped slightly in the comparison test. The reason was that there was a sparse distribution of sentence samples in the related category of comparative relations. In the DLOU-FSI corpus, each standard text of fishery contained only 0-3 comparison relation triples, less than 1% of the total number of all relational triples. As such, the comprehensive relationship features were necessary to learn for the higher recognition in the relationship extraction task. In addition, the quality of corpus was determined on the extraction of fishery standard text relations. The higher quality and the more accurate recognition of the model were achieved, as the deep learning model was trained to learn more data. There was also a critical value, after which the model cannot be greatly improved. In the case of a model adaptation, the test data can be effectively expanded to increase the number and diversity of samples for a better overall effect of relationship extraction. This finding can lay a significant foundation for the construction of fishery standard knowledge graphs. [ABSTRACT FROM AUTHOR]

Multi-dimensional classification (MDC) task can be considered the most inclusive description of all classifications tasks, as it joins multiple class spaces and their multiple class members into a single compound classification problem. The challenges in MDC arise from the possible class dependencies across different class spaces, as well as the imbalance of labels in training datasets due to lack of all possible combinations. In this paper, we propose a straightforward, yet efficient, MDC deep learning classifier, named Deep Self-Organizing Cube (DSOC) that can model dependencies among classes in multiple class spaces, while consolidating its ability to classify rare combinations of labels. DSOC is formed of two n -dimensional components, namely the Hypercube Classifier and the multiple DSOC Neural Networks connected to the hypercube. The multiple neural networks component is responsible for feature selection and segregation of classes, while the Hypercube classifier is responsible for creating the semantics among multiple class spaces and accommodate the model for rare sample classification. DSOC is a multiple-output learning algorithm that successfully classify samples across all class spaces simultaneously. To challenge the proposed DSOC model, we conducted an assessment on seventeen benchmark datasets in the four types of classification tasks, binary, multi-class, multi-label and multi-dimensional. The obtained results were compared to four standard classifiers and eight competitive state-of-the-art approaches reported in literature. The DSOC has achieved superior performance over standard classifiers as well as the state-of-the-art approaches in all the four classification tasks. Moreover, in terms of Exact Match accuracy metrics, DSOC has outperformed all state-of-the-art approaches in 77.8% of the cases, which reflects the superior ability of DSOC to model dependencies and successfully classify rare samples across all dimensions simultaneously. [ABSTRACT FROM AUTHOR]

The quantity of data generated increases daily, which makes it difficult to process. In the case of supervised learning, labeling training examples may represent an especially tedious and costly task. One of the aims of positive and unlabeled (PU) learning is to train a binary classifier from partially labeled data, representing a strategy for combining supervised and semi-supervised learning and alleviating the cost of labeling data fully. Still, the main strength of PU learning arises when the negative data are not directly available or too diverse. Although the generative approaches have shown promising results in this field, they also bring shortcomings, such as high computational cost, training instability, and inability to generate fully labeled datasets. In the paper, we propose a novel Conditional Generative PU framework (CGenPU) with a built-in auxiliary classifier. We develop a novel loss function to learn the distribution of positive and negative examples, which leads to a unique, desirable equilibrium under a nonparametric assumption. Our CGenPU is evaluated against existing generative approaches using both synthetic and real data. The characteristics of various methods, including ours, are depicted with different toy examples. The results demonstrate the state-of-the-art performance on standard positive and unlabeled learning benchmark datasets. Given only ten labeled CIFAR-10 examples, CGenPU achieves classification accuracy of 84%, while current state-of-the-art D-GAN framework achieves 54%. On top of that, CGenPU is the first single-stage generative framework for PU learning. • CGenPU is capable of conditional generation and binary classification. • Auxiliary PU loss enables the generation of positive and negative data distributions. • The objective has a unique, desirable equilibrium under a non-parametric assumption. • CGenPU shows state-of-the-art results in One-vs-One and One-vs-Rest scenarios. [ABSTRACT FROM AUTHOR]

Service composition can quickly build new value-added composite services by combining existing Web services. However, with the complex and changeable Internet environment, composite services need to effectively understand and manage requirements in a flexible and adaptive manner, and also be able to quickly and proactively provide high-quality services through a series of dynamic evolution models. In response to these challenges, in this study, we propose a service composition evolution method that can adaptively select the appropriate service to improve user satisfaction and service quality. First, we introduce a deep clustering method based on the topic model and auto-encoder, which considers the function description documents and parameter information of the services to reduce the search space of the candidate services. Furthermore, we describe a requirement-oriented context-sensitive task model to integrate functional and non-functional user requirements by connecting the context of Web subtasks. Then, we present a dynamic service matching method with QoS threshold judgement to filter and rank the services. We applied the proposed evolution method to the datasets of real-world Web services and measured the performance using standard measurement metrics. The prototype implementation and results of simulation experiments verified the effectiveness of each part of our method. • A novel service composition evolution method is proposed. • A deep clustering method is presented to implement the service clustering process. • RCT model and various evolutionary types are used to satisfy user requirements. • Simulation experiments on prototype system show the effectiveness. [ABSTRACT FROM AUTHOR]

Ideally, Convolutional Neural Networks (CNNs) should be trained with high quality images with minimum noise and correct ground truth labels. Nonetheless, in many real-world scenarios, such high quality is very hard to obtain, and datasets may be affected by any sort of image degradation and mislabelling issues. This negatively impacts the performance of standard CNNs, both during the training and the inference phase. To address this issue we propose Wise2WipedNet (W2WNet), a new two-module Convolutional Neural Network, where a Wise module exploits Bayesian inference to identify and discard spurious images during the training and a Wiped module takes care of the final classification, while broadcasting information on the prediction confidence at inference time. The goodness of our solution is demonstrated on a number of public benchmarks addressing different image classification tasks, as well as on a real-world case study on histological image analysis. Overall, our experiments demonstrate that W2WNet is able to identify image degradation and mislabelling issues both at training and at inference time, with positive impact on the final classification accuracy. • A novel two-module CNN exploiting Bayesian inference is introduced. • We address image degradation and mislabelling both at training and at inference. • Our method overcomes the deterministic counterpart (standard DenseNet121). • We show results on two public benchmarks and a real-world histological case study. [ABSTRACT FROM AUTHOR]

Microsatellite instability (MSI)/defective DNA mismatch repair (dMMR) is receiving more attention as a biomarker for eligibility for immune checkpoint inhibitors in advanced diseases. However, due to high costs and resource limitations, MSI/dMMR testing is not widely performed. Some attempts are in progress to predict MSI/dMMR status through histomorphological features on H&E slides using artificial intelligence (AI) technology. In this study, the potential predictive role of this new methodology was reviewed through a systematic review. Studies up to September 2021 were searched through PubMed and Embase database searches. The design and results of each study were summarized, and the risk of bias for each study was evaluated. For colorectal cancer, AI-based systems showed excellent performance with the highest standard of 0.972; for gastric and endometrial cancers they showed a relatively low but satisfactory performance, with the highest standard of 0.81 and 0.82, respectively. However, analyzing the risk of bias, most studies were evaluated at high-risk. AI-based systems showed a high potential in predicting the MSI/dMMR status of different cancer types, and particularly of colorectal cancers. Therefore, a confirmation test should be required only for the results that are positive in the AI test. [ABSTRACT FROM AUTHOR]

Bayesian computation plays an important role in modern machine learning and statistics to reason about uncertainty. A key computational challenge in Bayesian inference is to develop efficient techniques to approximate, or draw samples from posterior distributions. Stein variational gradient decent (SVGD) has been shown to be a powerful approximate inference algorithm for this issue. However, the vanilla SVGD requires calculating the gradient of the target density and cannot be applied when the gradient is unavailable or too expensive to evaluate. In this paper we explore one way to address this challenge by the construction of a local surrogate for the target distribution in which the gradient can be obtained in a much more computationally feasible manner. More specifically, we approximate the forward model using a deep neural network (DNN) which is trained on a carefully chosen training set, which also determines the quality of the surrogate. To this end, we propose a general adaptation procedure to refine the local approximation online without destroying the convergence of the resulting SVGD. This significantly reduces the computational cost of SVGD and leads to a suite of algorithms that are straightforward to implement. The new algorithm is illustrated on a set of challenging Bayesian inverse problems, and numerical experiments demonstrate a clear improvement in performance and applicability of standard SVGD. [ABSTRACT FROM AUTHOR]