Количество кликов
время
Количество загрузок
Классификация： 物理学 >> 核物理学 Время подачи： 2025-06-24
Wang, Mr. Haochen
Zhang, Mr. Zhao
Yu, Dr. Pei
Bao, Miss Yuxin
Zhai, Prof. Jiajia
Xu, Dr. Yu
Deng, Dr. Li
Xiao, Dr. Sa
Zhang, Dr. Xueheng
Yu, Dr. Yuhong
He, Dr. Weibo
Chen, Dr. Liangwen
Zhang, Prof. Yu
Yang, Prof. Lei 杨磊
Sun, Prof. Zhiyu 孙志宇
Аннотация： Cosmic-ray muon sources exhibit distinct scattering angle distributions when interacting with materials of different atomic numbers (Z values), facilitating the identification of various Z-class materials, particularly those radioactive high-Z nuclear elements. Most of the traditional identification methods are based on complex muon event reconstruction and trajectory fitting processes. Supervised machine learning methods offer some improvement but rely heavily on prior knowledge of target materials, significantly limiting their practical applicability in detecting concealed materials. For the first time, transfer learning is introduced into the field of muon tomography in this work. We propose two lightweight neural network models for fine-tuning and adversarial transfer learning, utilizing muon tomography data of bare materials to predict the Z-class of coated materials. By employing the inverse cumulative distribution function method, more accurate scattering angle distributions could be obtained from limited data, leading to an improvement by nearly 4% in prediction accuracy compared with the traditional random sampling based training. When applied to coated materials with limited labeled or even unlabeled muon tomography data, the proposed method achieves an overall prediction accuracy exceeding 96%, with high-Z materials reaching nearly 99%. Simulation results indicate that transfer learning improves prediction accuracy by approximately 10% compared to direct prediction without transfer. This study demonstrates the effectiveness of transfer learning in overcoming the physical challenges associated with limited labeled/unlabeled data, highlights the promising potential of transfer learning in the field of muon tomography.
Классификация： 物理学 >> 核物理学 Время подачи： 2025-05-28
Wang, Mr. Haochen
Zhang, Mr. Zhao
Yu, Dr. Pei
Bao, Miss Yuxin
Zhai, Prof. Jiajia
Xu, Dr. Yu
Deng, Dr. Li
Xiao, Dr. Sa
Zhang, Dr. Xueheng
Yu, Dr. Yuhong
He, Dr. Weibo
Chen, Dr. Liangwen
Zhang, Prof. Yu
Yang, Prof. Lei 杨磊
Sun, Prof. Zhiyu 孙志宇
Аннотация： Cosmic-ray muon sources exhibit distinct scattering angle distributions when interacting with materials of different atomic numbers (Z values), facilitating the identification of various Z-class materials, particularly those radioactive high-Z nuclear elements. Most of the traditional identification methods are based on complex muon event reconstruction and trajectory fitting processes. Supervised machine learning methods offer some improvement but rely heavily on prior knowledge of target materials, significantly limiting their practical applicability in detecting concealed materials. For the first time, transfer learning is introduced into the field of muon tomography in this work. We propose two lightweight neural network models for fine-tuning and adversarial transfer learning, utilizing muon tomography data of bare materials to predict the Z-class of coated materials. By employing the inverse cumulative distribution function method, more accurate scattering angle distributions could be obtained from limited data, leading to an improvement by nearly 4% in prediction accuracy compared with the traditional random sampling based training. When applied to coated materials with limited labeled or even unlabeled muon tomography data, the proposed method achieves an overall prediction accuracy exceeding 96%, with high-Z materials reaching nearly 99%. Simulation results indicate that transfer learning improves prediction accuracy by approximately 10% compared to direct prediction without transfer. This study demonstrates the effectiveness of transfer learning in overcoming the physical challenges associated with limited labeled/unlabeled data, highlights the promising potential of transfer learning in the field of muon tomography.
Классификация： 核科学技术 >> 辐射物理与技术 Время подачи： 2024-09-23
Аннотация： In scenarios such as vehicle radiation monitoring and unmanned aerial vehicle radiation detection, rapid measurements using a NaI(Tl) detector often result in low photon counts, weak characteristic peaks, and significant statistical fluctuations. These issues can lead to potential failures in peak-searching-based identification methods. To address the low precision associated with short-duration measurements of radionuclides, this paper proposes an identification algorithm that leverages heterogeneous spectral transfer to develop a low-count energy spectral identification model. Comparative experiments demonstrated that transferring samples from 26 classes of simulated heterogeneous gamma spectra aids in creating a reliable model for measured gamma spectra. With only 10% of target domain samples used for training, the accuracy on real low-count spectral samples was 95.56%. This performance shows a significant improvement over widely employed full-spectrum analysis methods trained on target domain samples. The proposed method also exhibits strong generalization capabilities, effectively mitigating overfitting issues in low-count energy spectral classification under short-duration measurements.
Классификация： 机械工程 >> 机械设计 Время подачи： 2024-04-01
Аннотация： Overstaffing production in underground coal mining is not convenient for daily management, and incomplete information of coal miners hinders the rescue process of firefighters during mine accidents. To address this safety sustainability issue, a novel face recognition method based on an improved multiscale neural network is proposed in this paper. A new depthwise seperable (DS)-inception block is designed and a joint supervised loss function based on center loss theory is developed to constructe a new multiscale model. The miniers can be recognized in the harsh underground environment during the life rescue. Experimental results show that the accuracy, recall and F1-score indexes of the proposed method for the miner face recognition in the underground mining environment are 97.26%, 94.17% and 95.42%, respectively. Transfer model with joint supervised loss can effectively improve the recognition accuracy by about 0.5~1.5%. In addition, the average recognition accuracy of the proposed face recognition method achieves to 91.34% and the miss detection rate is less than 5% in the dugout tunnel of coal mine.
Классификация： 计算机科学 >> 自然语言理解与机器翻译 Время подачи： 2019-05-12
Аннотация： Abstract. Computational chemistry develops fast in recent years due to the rapid growth and breakthroughs in AI. Thanks for the progress in natural language processing, researchers can extract more fine-grained knowledge in publications to stimulate the development in computational chemistry. While the works and corpora in chemical entity extraction have been restricted in the biomedicine or life science field instead of the chemistry field, we build a new corpus in chemical bond field anno- tated for 7 types of entities: compound, solvent, method, bond, reaction, pKa and pKa value. This paper presents a novel BERT-CRF model to build scientific chemical data chains by extracting 7 chemical entities and relations from publications. And we propose a joint model to ex- tract the entities and relations simultaneously. Experimental results on our Chemical Special Corpus demonstrate that we achieve state-of-art and competitive NER performance.
京ICP备05002861号-25 | 京公网安备110402500046号
版权所有© 2016 中国科学院文献情报中心
Будет рассмотрено
Количество кликов
Количество загрузок
комментарии