物理学报2025，Vol.74Issue(8)：41-50,10.DOI:10.7498/aps.74.20241775
基于变分自编码器的伽马单中子出射反应截面实验数据离群点研究
Outliers identification of experimental(γ,n)reaction cross section via variational autoencoder
摘要
Abstract
The(γ,n)cross-section is important in nuclear engineering transport calculations.The measurements of the(γ,n)reaction for some isotopes show significant discrepancies among different laboratories.Since the analysis of experimental data is the primary task in the evaluation of nuclear data,identifying the measured outlier data is crucial for improving the quality of nuclear data.Therefore,a variational autoencoder(VAE)method is used in this work to analyze experimental measurements of(γ,n)cross sections for nuclear mass ranging from 29 to 207,in order to provide more reliable experimental information for evaluating nuclear data. According to the proton number Z and nuclear mass A,we design a variational autoencoder network for outlier identification in the measurement of(γ,n).The silhouette coefficient method and K-means algorithm are used to cluster the latent variables of VAE.Subsequently,the experimental data with and without the outliers are compared with those from the IAEA-2019-PD to assess the VAE in its application to the evaluation of photoneutron measurements. The results demonstrate that the VAE can effectively identify outliers in the measurements of(γ,n).After excluding outliers,the(γ,n)cross-section for 54Fe,63Cu,181Ta,206 Pb,and 207Pb showed higher consistency with the IAEA-2019-PD evaluation results.However,29Si and 141Pr still deviate from the IAEA-2019-PD evaluation results,therefore requiring more analyses of the measurements themselves in future. The VAE method can effectively identify outliers and extract the latent structures in experimental data of(γ,n)reaction.It provides more reliable experimental information for evaluating nuclear data and validating the potential application of this method in nuclear data research.However,the generalizability of VAE method still needs further developing,especially in addressing the issues of uneven energy distribution for various measurements.关键词
变分自编码器/伽马单中子出射反应/截面/离群点Key words
variational autoencoder/(γ,n)reaction/cross section/outlier引用本文复制引用
谢金辰,陶曦,续瑞瑞,田源,邢康,葛智刚,牛一斐..基于变分自编码器的伽马单中子出射反应截面实验数据离群点研究[J].物理学报,2025,74(8):41-50,10.基金项目
国家重点研发计划(批准号:2022YFA1602403,2021YFA1601500)、国家自然科学基金(批准号:12005303,12075104)、重点实验室重点项目(批准号:JCKY2023201C153-5)和持续支持基础科学研究项目(批准号:BJ010261223282)资助的课题. Project supported by the National Key Research and Development Program of China(Grant Nos.2022YFA1602403,2021YFA1601500),the National Natural Science Foundation of China(Grant Nos.12005303,12075104),the Key Laboratory Key Project,China(Grant No.JCKY2023201C153-5),and the Continuous Support for Basic Scientific Research Project,China(Grant No.BJ010261223282). （批准号:2022YFA1602403,2021YFA1601500）
