Pub Date : 2024-07-15DOI: 10.1007/s40747-024-01499-9
Ji Lin, Quanliang Liu
Many engineering problems are essentially expensive multi-/many-objective optimization problems, and surrogate-assisted evolutionary algorithms have gained widespread attention in dealing with them. As the objective dimension increases, the error of predicting solutions based on surrogate models accumulates. Existing algorithms do not have strong selection pressure in the candidate solution obtaining and adaptive sampling stages. These make the effectiveness and area of application of the algorithms unsatisfactory. Therefore, this paper proposes a two-risk archive algorithm, which contains a strategy for mining high-risk and low-risk archives and a four-state adaptive sampling criterion. In the candidate solution mining stage, two types of Kriging models are trained, then conservative optimization models and non-conservative optimization models are constructed for model searching, followed by archive selection to obtain more reliable two-risk archives. In the adaptive sampling stage, in order to improve the performance of the algorithms, the proposed criterion considers environmental assessment, demand assessment, and sampling, where the sampling approach involves the improvement of the comprehensive performance in reliable environments, convergence and diversity in controversial environments, and surrogate model uncertainty. Experimental results on numerous benchmark problems show that the proposed algorithm is far superior to seven state-of-the-art algorithms in terms of comprehensive performance.
{"title":"TRAA: a two-risk archive algorithm for expensive many-objective optimization","authors":"Ji Lin, Quanliang Liu","doi":"10.1007/s40747-024-01499-9","DOIUrl":"https://doi.org/10.1007/s40747-024-01499-9","url":null,"abstract":"<p>Many engineering problems are essentially expensive multi-/many-objective optimization problems, and surrogate-assisted evolutionary algorithms have gained widespread attention in dealing with them. As the objective dimension increases, the error of predicting solutions based on surrogate models accumulates. Existing algorithms do not have strong selection pressure in the candidate solution obtaining and adaptive sampling stages. These make the effectiveness and area of application of the algorithms unsatisfactory. Therefore, this paper proposes a two-risk archive algorithm, which contains a strategy for mining high-risk and low-risk archives and a four-state adaptive sampling criterion. In the candidate solution mining stage, two types of Kriging models are trained, then conservative optimization models and non-conservative optimization models are constructed for model searching, followed by archive selection to obtain more reliable two-risk archives. In the adaptive sampling stage, in order to improve the performance of the algorithms, the proposed criterion considers environmental assessment, demand assessment, and sampling, where the sampling approach involves the improvement of the comprehensive performance in reliable environments, convergence and diversity in controversial environments, and surrogate model uncertainty. Experimental results on numerous benchmark problems show that the proposed algorithm is far superior to seven state-of-the-art algorithms in terms of comprehensive performance.</p>","PeriodicalId":10524,"journal":{"name":"Complex & Intelligent Systems","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141618247","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-15DOI: 10.1007/s40747-024-01528-7
Zhenwu Wang, Shiqi Zhang, Yang Chen, Mengjie Han, Yang Zhou, Benting Wan
In this paper, we address the challenges of random label ordering and limited interpretability associated with Ensemble Classifier Chains (ECC) by introducing a novel ECC method, ECC-MOO&BN, which integrates Bayesian Networks (BN) and Multi-Objective Optimization (MOO). This approach is designed to concurrently overcome these ECC limitations. The ECC-MOO&BN method focuses on extracting diverse and interpretable label orderings for the ECC classifier. We initiated this process by employing mutual information to investigate label relationships and establish the initial structures of the BN. Subsequently, an enhanced NSGA-II algorithm was applied to develop a series of Directed Acyclic Graphs (DAGs) that effectively balance the likelihood and complexity of the BN structure. The rationale behind using the MOO method lies in its ability to optimize both complexity and likelihood simultaneously, which not only diversifies DAG generation but also helps avoid overfitting during the production of label orderings. The DAGs, once sorted topologically, yielded a series of label orderings, which were then seamlessly integrated into the ECC framework for addressing multi-label classification (MLC) problems. Experimental results show that when benchmarked against eleven leading-edge MLC algorithms, our proposed method achieves the highest average ranking across seven evaluation criteria on nine out of thirteen MLC datasets. The results of the Friedman test and Nemenyi test also indicate that the performance of the proposed method has a significant advantage compared to other algorithms.
{"title":"A novel bayesian network-based ensemble classifier chains for multi-label classification","authors":"Zhenwu Wang, Shiqi Zhang, Yang Chen, Mengjie Han, Yang Zhou, Benting Wan","doi":"10.1007/s40747-024-01528-7","DOIUrl":"https://doi.org/10.1007/s40747-024-01528-7","url":null,"abstract":"<p>In this paper, we address the challenges of random label ordering and limited interpretability associated with Ensemble Classifier Chains (ECC) by introducing a novel ECC method, ECC-MOO&BN, which integrates Bayesian Networks (BN) and Multi-Objective Optimization (MOO). This approach is designed to concurrently overcome these ECC limitations. The ECC-MOO&BN method focuses on extracting diverse and interpretable label orderings for the ECC classifier. We initiated this process by employing mutual information to investigate label relationships and establish the initial structures of the BN. Subsequently, an enhanced NSGA-II algorithm was applied to develop a series of Directed Acyclic Graphs (DAGs) that effectively balance the likelihood and complexity of the BN structure. The rationale behind using the MOO method lies in its ability to optimize both complexity and likelihood simultaneously, which not only diversifies DAG generation but also helps avoid overfitting during the production of label orderings. The DAGs, once sorted topologically, yielded a series of label orderings, which were then seamlessly integrated into the ECC framework for addressing multi-label classification (MLC) problems. Experimental results show that when benchmarked against eleven leading-edge MLC algorithms, our proposed method achieves the highest average ranking across seven evaluation criteria on nine out of thirteen MLC datasets. The results of the Friedman test and Nemenyi test also indicate that the performance of the proposed method has a significant advantage compared to other algorithms.</p>","PeriodicalId":10524,"journal":{"name":"Complex & Intelligent Systems","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141618249","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Hydrocracking represents a complex and time-consuming chemical process that converts heavy oil fractions into various valuable products with low boiling points. It plays a pivotal role in enhancing the quality of products within the oil refining process. Consequently, the development of efficient surrogate models for simulating the hydrocracking process and identifying appropriate solutions for multi-objective oil refining is now an important area of research. In this study, a novel transferable preference learning-driven evolutionary algorithm is proposed to facilitate multi-objective decision analysis in the oil refining process. Specifically, our approach involves considering user preferences to divide the objective space into a region of interest (ROI) and other subspaces. We then utilize Kriging models to approximate the sub-problems within the ROI. In order to enhance the robustness and generalization capability of the Kriging models during the evolutionary process, we transfer the mutual information between the sub-problems in the ROI. To validate the effectiveness as well as efficiency of our proposed method, we undertake a series of experiments on both benchmarks and the oil refining process. The experimental results conclusively demonstrate the superiority of our approach.
加氢裂化是一种复杂而耗时的化学工艺,可将重油馏分转化为各种有价值的低沸点产品。在炼油过程中,加氢裂化对提高产品质量起着至关重要的作用。因此,开发高效的替代模型来模拟加氢裂化过程并为多目标炼油确定合适的解决方案是目前的一个重要研究领域。本研究提出了一种新颖的可转移偏好学习驱动的进化算法,以促进炼油过程中的多目标决策分析。具体来说,我们的方法包括考虑用户偏好,将目标空间划分为感兴趣区域(ROI)和其他子空间。然后,我们利用克里金模型对 ROI 内的子问题进行近似。为了增强 Kriging 模型在演化过程中的稳健性和泛化能力,我们在 ROI 中转移了子问题之间的互信息。为了验证我们提出的方法的有效性和效率,我们在基准和炼油过程中进行了一系列实验。实验结果充分证明了我们方法的优越性。
{"title":"Transferable preference learning in multi-objective decision analysis and its application to hydrocracking","authors":"Guo Yu, Xinzhe Wang, Chao Jiang, Yang Liu, Lianbo Ma, Cuimei Bo, Quanling Zhang","doi":"10.1007/s40747-024-01537-6","DOIUrl":"https://doi.org/10.1007/s40747-024-01537-6","url":null,"abstract":"<p>Hydrocracking represents a complex and time-consuming chemical process that converts heavy oil fractions into various valuable products with low boiling points. It plays a pivotal role in enhancing the quality of products within the oil refining process. Consequently, the development of efficient surrogate models for simulating the hydrocracking process and identifying appropriate solutions for multi-objective oil refining is now an important area of research. In this study, a novel transferable preference learning-driven evolutionary algorithm is proposed to facilitate multi-objective decision analysis in the oil refining process. Specifically, our approach involves considering user preferences to divide the objective space into a region of interest (ROI) and other subspaces. We then utilize Kriging models to approximate the sub-problems within the ROI. In order to enhance the robustness and generalization capability of the Kriging models during the evolutionary process, we transfer the mutual information between the sub-problems in the ROI. To validate the effectiveness as well as efficiency of our proposed method, we undertake a series of experiments on both benchmarks and the oil refining process. The experimental results conclusively demonstrate the superiority of our approach.</p>","PeriodicalId":10524,"journal":{"name":"Complex & Intelligent Systems","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141618250","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-12DOI: 10.1007/s40747-024-01545-6
Richárd Kiss, Gábor Szűcs
Network science has witnessed a surge in popularity, driven by the transformative power of node representation learning for diverse applications like social network analysis and biological modeling. While shallow embedding algorithms excel at capturing network structure, they face a critical limitation—failing to generalize to unseen nodes. This paper addresses this challenge by introducing Inductive Shallow Node Embedding—as a main contribution—pioneering a novel approach that extends shallow embeddings to the realm of inductive learning. It has a novel encoder architecture that captures the local neighborhood structure of each node, enabling effective generalization to unseen nodes. In the generalization, robustness is essential to avoid degradation of performance arising from noise in the dataset. It has been theoretically proven that the covariance of the additive noise term in the proposed model is inversely proportional to the cardinality of a node’s neighbors. Another contribution is a mathematical lower bound to quantify the robustness of node embeddings, confirming its advantage over traditional shallow embedding methods, particularly in the presence of parameter noise. The proposed method demonstrably excels in dynamic networks, consistently achieving over 90% performance on previously unseen nodes compared to nodes encountered during training on various benchmarks. The empirical evaluation concludes that our method outperforms competing methods on the vast majority of datasets in both transductive and inductive tasks.
{"title":"Unsupervised Graph Representation Learning with Inductive Shallow Node Embedding","authors":"Richárd Kiss, Gábor Szűcs","doi":"10.1007/s40747-024-01545-6","DOIUrl":"https://doi.org/10.1007/s40747-024-01545-6","url":null,"abstract":"<p>Network science has witnessed a surge in popularity, driven by the transformative power of node representation learning for diverse applications like social network analysis and biological modeling. While shallow embedding algorithms excel at capturing network structure, they face a critical limitation—failing to generalize to unseen nodes. This paper addresses this challenge by introducing Inductive Shallow Node Embedding—as a main contribution—pioneering a novel approach that extends shallow embeddings to the realm of inductive learning. It has a novel encoder architecture that captures the local neighborhood structure of each node, enabling effective generalization to unseen nodes. In the generalization, robustness is essential to avoid degradation of performance arising from noise in the dataset. It has been theoretically proven that the covariance of the additive noise term in the proposed model is inversely proportional to the cardinality of a node’s neighbors. Another contribution is a mathematical lower bound to quantify the robustness of node embeddings, confirming its advantage over traditional shallow embedding methods, particularly in the presence of parameter noise. The proposed method demonstrably excels in dynamic networks, consistently achieving over 90% performance on previously unseen nodes compared to nodes encountered during training on various benchmarks. The empirical evaluation concludes that our method outperforms competing methods on the vast majority of datasets in both transductive and inductive tasks.</p>","PeriodicalId":10524,"journal":{"name":"Complex & Intelligent Systems","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141597682","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Tracking is a crucial problem for nonlinear systems as it ensures stability and enables the system to accurately follow a desired reference signal. Using Takagi–Sugeno (T–S) fuzzy models, this paper addresses the problem of fuzzy observer and control design for a class of nonlinear systems. The Takagi–Sugeno (T–S) fuzzy models can represent nonlinear systems because it is a universal approximation. Firstly, the T–S fuzzy modeling is applied to get the dynamics of an observational system in order to estimate the unmeasurable states of an unknown nonlinear system. There are various kinds of nonlinear systems that can be modeled using T–S fuzzy systems by combining the input state variables linearly. Secondly, the T–S fuzzy systems can handle unknown states as well as parameters known to the indirect adaptive fuzzy observer. A simple feedback method is used to implement the proposed controller. As a result, the feedback linearization method allows for solving the singularity problem without using any additional algorithms. A fuzzy model representation of the observation system comprises parameters and a feedback gain. The Lyapunov function and Lipschitz conditions are used in constructing the adaptive law. This method is then illustrated by an illustrative example to prove its effectiveness with different kinds of nonlinear functions. A well-designed controller is effective and its performance index minimizes network utilization—this factor is particularly significant when applied to wireless communication systems.
{"title":"Indirect adaptive observer control (I-AOC) design for truck–trailer model based on T–S fuzzy system with unknown nonlinear function","authors":"Muhammad Shamrooz Aslam, Hazrat Bilal, Wer-jer Chang, Abid Yahya, Irfan Anjum Badruddin, Sarfaraz Kamangar, Mohamed Hussien","doi":"10.1007/s40747-024-01544-7","DOIUrl":"https://doi.org/10.1007/s40747-024-01544-7","url":null,"abstract":"<p>Tracking is a crucial problem for nonlinear systems as it ensures stability and enables the system to accurately follow a desired reference signal. Using Takagi–Sugeno (T–S) fuzzy models, this paper addresses the problem of fuzzy observer and control design for a class of nonlinear systems. The Takagi–Sugeno (T–S) fuzzy models can represent nonlinear systems because it is a universal approximation. Firstly, the T–S fuzzy modeling is applied to get the dynamics of an observational system in order to estimate the unmeasurable states of an unknown nonlinear system. There are various kinds of nonlinear systems that can be modeled using T–S fuzzy systems by combining the input state variables linearly. Secondly, the T–S fuzzy systems can handle unknown states as well as parameters known to the indirect adaptive fuzzy observer. A simple feedback method is used to implement the proposed controller. As a result, the feedback linearization method allows for solving the singularity problem without using any additional algorithms. A fuzzy model representation of the observation system comprises parameters and a feedback gain. The Lyapunov function and Lipschitz conditions are used in constructing the adaptive law. This method is then illustrated by an illustrative example to prove its effectiveness with different kinds of nonlinear functions. A well-designed controller is effective and its performance index minimizes network utilization—this factor is particularly significant when applied to wireless communication systems.</p>","PeriodicalId":10524,"journal":{"name":"Complex & Intelligent Systems","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141597692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Policy Gradient (PG) method is one of the most popular algorithms in Reinforcement Learning (RL). However, distributed adaptive variants of PG are rarely studied in multi-agent. For this reason, this paper proposes a distributed adaptive policy gradient algorithm (IS-DAPGM) incorporated with Adam-type updates and importance sampling technique. Furthermore, we also establish the theoretical convergence rate of (mathcal {O}(1/sqrt{T})), where T represents the number of iterations, it can match the convergence rate of the state-of-the-art centralized policy gradient methods. In addition, many experiments are conducted in a multi-agent environment, which is a modification on the basis of Particle world environment. By comparing with some other distributed PG methods and changing the number of agents, we verify the performance of IS-DAPGM is more efficient than the existing methods.
策略梯度法(PG)是强化学习(RL)中最流行的算法之一。然而,PG 的分布式自适应变体在多智能体中很少被研究。为此,本文提出了一种分布式自适应策略梯度算法(IS-DAPGM),该算法结合了亚当型更新和重要性采样技术。此外,我们还建立了理论收敛速率为(1//sqrt{T})的分布式自适应策略梯度算法(IS-DAPGM),其中 T 代表迭代次数,它可以与最先进的集中式策略梯度方法的收敛速率相媲美。此外,许多实验都是在多代理环境下进行的,这是在粒子世界环境基础上进行的修改。通过与其他一些分布式策略梯度方法的比较以及改变代理数量,我们验证了 IS-DAPGM 的性能比现有方法更高效。
{"title":"A distributed adaptive policy gradient method based on momentum for multi-agent reinforcement learning","authors":"Junru Shi, Xin Wang, Mingchuan Zhang, Muhua Liu, Junlong Zhu, Qingtao Wu","doi":"10.1007/s40747-024-01529-6","DOIUrl":"https://doi.org/10.1007/s40747-024-01529-6","url":null,"abstract":"<p>Policy Gradient (PG) method is one of the most popular algorithms in Reinforcement Learning (RL). However, distributed adaptive variants of PG are rarely studied in multi-agent. For this reason, this paper proposes a distributed adaptive policy gradient algorithm (IS-DAPGM) incorporated with Adam-type updates and importance sampling technique. Furthermore, we also establish the theoretical convergence rate of <span>(mathcal {O}(1/sqrt{T}))</span>, where <i>T</i> represents the number of iterations, it can match the convergence rate of the state-of-the-art centralized policy gradient methods. In addition, many experiments are conducted in a multi-agent environment, which is a modification on the basis of Particle world environment. By comparing with some other distributed PG methods and changing the number of agents, we verify the performance of IS-DAPGM is more efficient than the existing methods.</p>","PeriodicalId":10524,"journal":{"name":"Complex & Intelligent Systems","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141597678","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-11DOI: 10.1007/s40747-024-01532-x
Ghada M. El-Banby, Nourhan S. Salem, Eman A. Tafweek, Essam N. Abd El-Azziz
Breast cancer is the second most prevalent cause of cancer death and the most common malignancy among women, posing a life-threatening risk. Treatment for breast cancer can be highly effective, with a survival chance of 90% or higher, especially when the disease is detected early. This paper introduces a groundbreaking deep U-Net framework for mammography breast cancer images to perform automatic detection of abnormalities. The objective is to provide segmented images that show areas of tumors more accurately than other deep learning techniques. The proposed framework consists of three steps. The first step is image preprocessing using the Li algorithm to minimize the cross-entropy between the foreground and the background, contrast enhancement using contrast-limited adaptive histogram equalization (CLAHE), normalization, and median filtering. The second step involves data augmentation to mitigate overfitting and underfitting, and the final step is implementing a convolutional encoder-decoder network-based U-Net architecture, characterized by high precision in medical image analysis. The framework has been tested on two comprehensive public datasets, namely INbreast and CBIS-DDSM. Several metrics have been adopted for quantitative performance assessment, including the Dice score, sensitivity, Hausdorff distance, Jaccard coefficient, precision, and F1 score. Quantitative results on the INbreast dataset show an average Dice score of 85.61% and a sensitivity of 81.26%. On the CBIS-DDSM dataset, the average Dice score is 87.98%, and the sensitivity reaches 90.58%. The experimental results ensure earlier and more accurate abnormality detection. Furthermore, the success of the proposed deep learning framework in mammography shows promise for broader applications in medical imaging, potentially revolutionizing various radiological practices.
乳腺癌是第二大癌症死因,也是女性最常见的恶性肿瘤,对生命构成威胁。乳腺癌的治疗效果显著,尤其是在早期发现乳腺癌时,患者的生存率可达 90% 或更高。本文介绍了一种开创性的深度 U-Net 框架,可用于乳房 X 射线照相术乳腺癌图像的异常自动检测。其目标是提供比其他深度学习技术更准确地显示肿瘤区域的分割图像。所提出的框架包括三个步骤。第一步是使用 Li 算法对图像进行预处理,使前景与背景之间的交叉熵最小化;使用对比度限制自适应直方图均衡化(CLAHE)增强对比度;归一化和中值滤波。第二步涉及数据增强,以减轻过拟合和欠拟合,最后一步是实施基于卷积编码器-解码器网络的 U-Net 架构,该架构在医学图像分析中具有高精度的特点。该框架已在两个综合公共数据集(即 INbreast 和 CBIS-DDSM)上进行了测试。定量性能评估采用了多个指标,包括 Dice 分数、灵敏度、Hausdorff 距离、Jaccard 系数、精确度和 F1 分数。INbreast 数据集的定量结果显示,平均 Dice 得分为 85.61%,灵敏度为 81.26%。在 CBIS-DDSM 数据集上,平均 Dice 得分为 87.98%,灵敏度达到 90.58%。实验结果确保了更早更准确地检测到异常。此外,所提出的深度学习框架在乳腺 X 射线照相术中的成功应用为医学影像领域的更广泛应用带来了希望,有可能彻底改变各种放射学实践。
{"title":"Automated abnormalities detection in mammography using deep learning","authors":"Ghada M. El-Banby, Nourhan S. Salem, Eman A. Tafweek, Essam N. Abd El-Azziz","doi":"10.1007/s40747-024-01532-x","DOIUrl":"https://doi.org/10.1007/s40747-024-01532-x","url":null,"abstract":"<p>Breast cancer is the second most prevalent cause of cancer death and the most common malignancy among women, posing a life-threatening risk. Treatment for breast cancer can be highly effective, with a survival chance of 90% or higher, especially when the disease is detected early. This paper introduces a groundbreaking deep U-Net framework for mammography breast cancer images to perform automatic detection of abnormalities. The objective is to provide segmented images that show areas of tumors more accurately than other deep learning techniques. The proposed framework consists of three steps. The first step is image preprocessing using the Li algorithm to minimize the cross-entropy between the foreground and the background, contrast enhancement using contrast-limited adaptive histogram equalization (CLAHE), normalization, and median filtering. The second step involves data augmentation to mitigate overfitting and underfitting, and the final step is implementing a convolutional encoder-decoder network-based U-Net architecture, characterized by high precision in medical image analysis. The framework has been tested on two comprehensive public datasets, namely INbreast and CBIS-DDSM. Several metrics have been adopted for quantitative performance assessment, including the Dice score, sensitivity, Hausdorff distance, Jaccard coefficient, precision, and F1 score. Quantitative results on the INbreast dataset show an average Dice score of 85.61% and a sensitivity of 81.26%. On the CBIS-DDSM dataset, the average Dice score is 87.98%, and the sensitivity reaches 90.58%. The experimental results ensure earlier and more accurate abnormality detection. Furthermore, the success of the proposed deep learning framework in mammography shows promise for broader applications in medical imaging, potentially revolutionizing various radiological practices.</p>","PeriodicalId":10524,"journal":{"name":"Complex & Intelligent Systems","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141584530","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-10DOI: 10.1007/s40747-024-01539-4
Leilei Cao, Yibo Guo, Ye Yuan, Qiangguo Jin
Few-shot Semantic Segmentation (FSS) was proposed to segment unseen classes in a query image, referring to only a few annotated examples named support images. One of the characteristics of FSS is spatial inconsistency between query and support targets, e.g., texture or appearance. This greatly challenges the generalization ability of methods for FSS, which requires to effectively exploit the dependency of the query image and the support examples. Most existing methods abstracted support features into prototype vectors and implemented the interaction with query features using cosine similarity or feature concatenation. However, this simple interaction may not capture spatial details in query features. To address this limitation, some methods utilized pixel-level support information by computing pixel-level correlations between paired query and support features implemented with the attention mechanism of Transformer. Nevertheless, these approaches suffer from heavy computation due to dot-product attention between all pixels of support and query features. In this paper, we propose a novel framework, termed ProtoFormer, built upon the Transformer architecture, to fully capture spatial details in query features. ProtoFormer treats the abstracted prototype of the target class in support features as the Query and the query features as Key and Value embeddings, which are input to the Transformer decoder. This approach enables better capture of spatial details and focuses on the semantic features of the target class in the query image. The output of the Transformer-based module can be interpreted as semantic-aware dynamic kernels that filter the segmentation mask from the enriched query features. Extensive experiments conducted on PASCAL-(5^{i}) and COCO-(20^{i}) datasets demonstrate that ProtoFormer significantly outperforms the state-of-the-art methods in FSS.
{"title":"Prototype as query for few shot semantic segmentation","authors":"Leilei Cao, Yibo Guo, Ye Yuan, Qiangguo Jin","doi":"10.1007/s40747-024-01539-4","DOIUrl":"https://doi.org/10.1007/s40747-024-01539-4","url":null,"abstract":"<p>Few-shot Semantic Segmentation (FSS) was proposed to segment unseen classes in a query image, referring to only a few annotated examples named support images. One of the characteristics of FSS is spatial inconsistency between query and support targets, e.g., texture or appearance. This greatly challenges the generalization ability of methods for FSS, which requires to effectively exploit the dependency of the query image and the support examples. Most existing methods abstracted support features into prototype vectors and implemented the interaction with query features using cosine similarity or feature concatenation. However, this simple interaction may not capture spatial details in query features. To address this limitation, some methods utilized pixel-level support information by computing pixel-level correlations between paired query and support features implemented with the attention mechanism of Transformer. Nevertheless, these approaches suffer from heavy computation due to dot-product attention between all pixels of support and query features. In this paper, we propose a novel framework, termed ProtoFormer, built upon the Transformer architecture, to fully capture spatial details in query features. ProtoFormer treats the abstracted prototype of the target class in support features as the Query and the query features as Key and Value embeddings, which are input to the Transformer decoder. This approach enables better capture of spatial details and focuses on the semantic features of the target class in the query image. The output of the Transformer-based module can be interpreted as semantic-aware dynamic kernels that filter the segmentation mask from the enriched query features. Extensive experiments conducted on PASCAL-<span>(5^{i})</span> and COCO-<span>(20^{i})</span> datasets demonstrate that ProtoFormer significantly outperforms the state-of-the-art methods in FSS.</p>","PeriodicalId":10524,"journal":{"name":"Complex & Intelligent Systems","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141566310","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-09DOI: 10.1007/s40747-024-01538-5
Fen Zhao, Huishuang Shao, Shuo Li, Yintong Wang, Yan Yu
Natural answer generation is in a very clear practical significance and strong application background, which can be widely used in the field of knowledge services such as community question answering and intelligent customer service. Traditional knowledge question answering is to provide precise answer entities and neglect the defects; namely, users hope to receive a complete natural answer. In this research, we propose a novel attention-based recurrent neural network for natural answer generation, which is enhanced with multi-level copying mechanisms and question-aware loss. To generate natural answers that conform to grammar, we leverage multi-level copying mechanisms and the prediction mechanism which can copy semantic units and predict common words. Moreover, considering the problem that the generated natural answer does not match the user question, question-aware loss is introduced to make the generated target answer sequences correspond to the question. Experiments on three response generation tasks show our model to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 0.727 BLEU on the SimpleQuestions response generation task, improving over the existing best results by over 0.007 BLEU. Our model has scored a significant enhancement on naturalness with up to 0.05 more than best performing baseline. The simulation results show that our method can generate grammatical and contextual natural answers according to user needs.
{"title":"Attention-based RNN with question-aware loss and multi-level copying mechanism for natural answer generation","authors":"Fen Zhao, Huishuang Shao, Shuo Li, Yintong Wang, Yan Yu","doi":"10.1007/s40747-024-01538-5","DOIUrl":"https://doi.org/10.1007/s40747-024-01538-5","url":null,"abstract":"<p>Natural answer generation is in a very clear practical significance and strong application background, which can be widely used in the field of knowledge services such as community question answering and intelligent customer service. Traditional knowledge question answering is to provide precise answer entities and neglect the defects; namely, users hope to receive a complete natural answer. In this research, we propose a novel attention-based recurrent neural network for natural answer generation, which is enhanced with multi-level copying mechanisms and question-aware loss. To generate natural answers that conform to grammar, we leverage multi-level copying mechanisms and the prediction mechanism which can copy semantic units and predict common words. Moreover, considering the problem that the generated natural answer does not match the user question, question-aware loss is introduced to make the generated target answer sequences correspond to the question. Experiments on three response generation tasks show our model to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 0.727 BLEU on the SimpleQuestions response generation task, improving over the existing best results by over 0.007 BLEU. Our model has scored a significant enhancement on naturalness with up to 0.05 more than best performing baseline. The simulation results show that our method can generate grammatical and contextual natural answers according to user needs.</p>","PeriodicalId":10524,"journal":{"name":"Complex & Intelligent Systems","volume":null,"pages":null},"PeriodicalIF":5.8,"publicationDate":"2024-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141561358","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}