In order to explore the correlation between different MRI sequences and the results of U-Net segmentation of glioma subregions, this paper proposes an interpretable method based on an evolutionary integration algorith...
详细信息
ISBN:
(数字)9798350349115
ISBN:
(纸本)9798350349122
In order to explore the correlation between different MRI sequences and the results of U-Net segmentation of glioma subregions, this paper proposes an interpretable method based on an evolutionary integration algorithm for logical discovery of the segmentation process of U-Net. Our approach consists of three steps: 1) Global fitting of the U-Net model to the segmentation results of gliomas using a dual evolutionary algorithm to generate a fitted model with both accuracy and interpretability.2) Extracting decision rules from the fitted model according to a specific target interpretable region and generating a complete set of interpretable rules after optimisation.3) Proposing a decision path integrator modeling method for the target region of decision paths for experimental validation. In this study, 293 patients from the BraTS2020 dataset are used as research data, and the accuracy of the fitted model is obtained to be 0.92, which is basically the same as that of Random Forest, but the model in this study has a better and simpler internal structure. At the same time, this study validated the relationship between Flair sequence and the edema region of glioma, and the experimental results showed that our extracted decision paths have a certain auxiliary effect on the segmentation of U-Net, and also proved the effectiveness of our proposed interpretability method.
3D hand pose for a single depth image is an important topic in computer vision and human-computer interaction, and although significant progress has been made in this field in recent years, there is still room for imp...
详细信息
ISBN:
(数字)9798350349115
ISBN:
(纸本)9798350349122
3D hand pose for a single depth image is an important topic in computer vision and human-computer interaction, and although significant progress has been made in this field in recent years, there is still room for improvement in accuracy for some specific application scenarios. To address this problem, a 3D hand pose estimation algorithm based on external attention is proposed. First, the target features are extracted by an hourglass network; then, a HEM (Hard Example Mining) loss based on a mean-variance loss function is proposed, which firstly calculates the L2 loss values of all N keypoints, and then sorts these loss values, and back-propagates the gradient only to the first m loss values. Meanwhile, external attention is introduced to enhance the ability to perceive the global information of the target, and the recognition ability of the features is improved by giving the features different influences through the attention weights. Experimental results show that the algorithm achieves an average distance error of 5.42 mm on the ICVL dataset and 7.11 mm on the MSRA dataset, which further improves the detection performance of the 3D hand pose estimation algorithm.
Multivariate time series forecasting is particularly critical in a number of areas involving air forecasting, electricity consumption and exchange rate transformations. The time series data in these fields often conta...
详细信息
ISBN:
(数字)9798350349115
ISBN:
(纸本)9798350349122
Multivariate time series forecasting is particularly critical in a number of areas involving air forecasting, electricity consumption and exchange rate transformations. The time series data in these fields often contain present multivariate nature, fusing long and short-term patterns, and traditional forecasting methods are often difficult to work. To address this challenge, this paper proposes a new deep learning framework, TSARNet, which combines gated recurrent units (GRUs) and an adaptive scale processing layer to effectively capture short-term localized dependency patterns among multivariate variables and reveal long-term trends through a spatio-temporal attention mechanism. In addition, an autoregressive model is incorporated to enhance the network's sensitivity to time series scale changes. Comprehensive experiments on four publicly available datasets show that the proposed TSARNet algorithm outperforms other comparative methods in most cases in terms of mean absolute error (MAE) and root mean square error (RMSE) in the long term prediction of different datasets. The TSARNet method also achieves optimal results for doing long-term prediction for each dimension of the power transformer temperature dataset.
The challenge of solving dynamic multi-objective optimization problems is to trace the varying Pareto optimal front and/or Pareto optimal set quickly and efficiently. This paper proposes a multi-direction prediction s...
详细信息
In order to solve the impact of the temporal and spatial characteristics of traffic on network routing optimization, this paper proposes convolution long-short memory neural network deep reinforcement learning (CLSDRL...
详细信息
Industrial big data was usually multi-source, heterogeneous, and deeply intertwined. It had a wide range of data sources, high data dimensions, and strong data correlation. In order to effectively analyze and process ...
详细信息
Industrial big data was usually multi-source, heterogeneous, and deeply intertwined. It had a wide range of data sources, high data dimensions, and strong data correlation. In order to effectively analyze and process streaming industrial big data generated by edge computing, it was very important to provide an effective real-time incremental data method. However, in the process of incremental processing, industrial big data incremental computing faced the challenges of dimensional disaster, repeated calculations, and the explosion of intermediate results. Therefore, in order to solve the above problems effectively, a QR-based tensor-train(TT) decomposition(TTD) method and a QR-based incremental TTD(QRITTD) method were proposed. This algorithm combined the incremental QR-based decomposition algorithm with an approximate singular value decomposition(SVD) algorithm and had good scalability. In addition, the computational complexity, space complexity, and approximation error analysis were analyzed in detail. The effectiveness of the three algorithms of QRITTD, non-incremental TTD(NITTD), and TT rank-1(TTr1) SVD(TTr1 SVD)were verified by comparison. Experimental results show that the SVD QRITTD method has better performance under the premise of ensuring the same tensor size.
Multifunctional therapeutic peptides(MFTP) hold immense potential in diverse therapeutic contexts, yet their prediction and identification remain challenging due to the limitations of traditional methodologies, such a...
Multifunctional therapeutic peptides(MFTP) hold immense potential in diverse therapeutic contexts, yet their prediction and identification remain challenging due to the limitations of traditional methodologies, such as extensive training durations, limited sample sizes, and inadequate generalization capabilities. To address these issues, we present AMHF-TP, an advanced method for MFTP recognition that utilizes attention mechanisms and multi-granularity hierarchical features to enhance performance. The AMHF-TP is composed of four key components: a migration learning module that leverages pretrained models to extract atomic compositional features of MFTP sequences; a convolutional neural network and selfattention module that refine feature extraction from amino acid sequences and their secondary structures; a hypergraph module that constructs a hypergraph for complex similarity representation between MFTP sequences; and a hierarchical feature extraction module that integrates multimodal peptide sequence features. Compared with leading methods,the proposed AMHF-TP demonstrates superior precision, accuracy, and coverage, underscoring its effectiveness and robustness in MFTP recognition. The comparative analysis of separate hierarchical models and the combined model, as well as with five contemporary models, reveals AMHF-TP's exceptional performance and stability in recognition tasks.
Context: Reliable and effective similarity analysis for the smart contracts facilitates the maintenance and quality assurance of the smart contract ecosystem. However, existing signature-based methods and code represe...
详细信息
Context: Reliable and effective similarity analysis for the smart contracts facilitates the maintenance and quality assurance of the smart contract ecosystem. However, existing signature-based methods and code representation learning-based methods suffer from limitations such as heavy-weight program analysis payloads or suboptimal contract encodings. Objective: This paper aims to design a fully unsupervised language model for better capturing the syntactic and semantic richness of Solidity code, and utilizes it for advancing the effectiveness of smart contract similarity analysis. Methods: Inspired by the impressive semantic learning capability of pre-trained language models (PLMs), we propose SolBERT, a PLM specifically tailored for enhancing Solidity smart contracts similarity detection. To ensure it produces high-quality encodings, SolBERT leverages BERT-style pre-training with the masked language modeling (MLM) and token type prediction (TTP) tasks applied on code-structure-aware token sequences derived from the contracts’ abstract syntax trees (ASTs) through structure-retaining tree linearization and light-weight normalization to learn a base model. On this basis, self-supervised contrastive fine-tuning and unsupervised whitening operations are further performed to optimize contract encoding generation. Results: Experiments are conducted on three contract similarity-related tasks, including contract clone detection, bug detection, and code clustering. The results indicate that SolBERT significantly outperforms state-of-the-art approaches with average absolute gains of 21.33% and 21.50% in terms of F1, and 17.78% and 26.60% in terms of accuracy for the clone detection and bug detection tasks, respectively;and an average absolute gain of 17.97% for code clustering task. When applying both contrastive fine-tuning and whitening optimizations, SolBERT also shows superior performance than the case of lacking any of them. Conclusion: The proposed approach, SolBERT, ca
Image shadow removal is challenging due to the diversity of shadows and the dependence on the background of the area covered by shadows. In this paper, we design an enhanced generative adversarial network (EGAN) to ha...
Image shadow removal is challenging due to the diversity of shadows and the dependence on the background of the area covered by shadows. In this paper, we design an enhanced generative adversarial network (EGAN) to handle the shadow removal task. EGAN is a complicated generative adversarial network that may exploit previous info, like shadow masks and lightness estimation, for improving recovery capability. In addition, EGAN can effectively utilize the details of shadow regions and eliminate artifacts. EGAN owns three main advantages. First, EGAN employs stricter rules for discriminators, which can upgrade the instability of traditional GAN. Second, EGAN involves a completely unique enhanced attention module, which may efficiently pick up different types of previous info to accelerate the learning efficiency of image reconstruction. Finally, EGAN is essentially an original cascaded network, that is, DRDU, which is beneficial for efficiently detecting hierarchical features of image recovery. Simulation experiments are conducted to verify the effectiveness of the proposed method on two benchmark datasets including SRD and ISTD, the strengths of our proposed method are indicated over the current representative ones according to subjective performance and objective estimation indexes.
To enhance the computational efficiency and precision of community discovery, a community discovery algorithm with the mixed label based on the minimum description length (MDI) of information compression is proposed i...
详细信息
暂无评论