Study of photoneutron reaction based on physics-informed Bayesian neural network

Qiankun SUN; Yue ZHANG; Zirui HAO; Hongwei WANG; Gongtao FAN; Hanghua XU; Longxiang LIU; Kaijie CHEN; Sheng JIN; Zhenwei WANG; Mengke XU; Xiangfei WANG

doi:10.11889/j.0253-3219.2025.hjs.48.240501

NUCLEAR TECHNIQUES, Volume. 48, Issue 5, 050001(2025)

Study of photoneutron reaction based on physics-informed Bayesian neural network

Qiankun SUN^1,2, Yue ZHANG³, Zirui HAO³, Hongwei WANG^1,2,3、*, Gongtao FAN^1,2,3, Hanghua XU³, Longxiang LIU³, Kaijie CHEN^1,4, Sheng JIN^1,2, Zhenwei WANG^1,2, Mengke XU^1,2, and Xiangfei WANG^1,2

Author Affiliations

¹Shanghai Institute of Applied Physics, Chinese Academy of Sciences, Shanghai 201800, China

²University of Chinese Academy of Sciences, Beijing 100049, China

³Shanghai Advanced Research Institute, Chinese Academy of Sciences, Shanghai 201210, China

⁴ShanghaiTech University, Shanghai 201210, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(7)

Fig. 1. Diagram of physics-informed Bayesian neural network for photoneutron reaction

Download full size

View in Article

Fig. 2. Cross-sections prediction of $(γ, n)$ and $(γ, 2 n)$ reaction channel of ${}^{89} Y$ in training set by Bayesian neural networks with different hidden layers, experimental data were taken from the Livermore, available in EXFOR database^[28]

Download full size

View in Article

Fig. 3. Cross-sections prediction of $(γ, n)$ and $(γ, 2 n)$ reaction channel of ${}^{159} T b$ in training set by Bayesian neural networks with different hidden layers, experimental data were taken from the Livermore, available in EXFOR database^[28]

Download full size

View in Article

Fig. 4. Cross-sections prediction of $(γ, n)$ and $(γ, 2 n)$ reaction channel of ${}^{197} A u$ by Bayesian neural networks with different hidden layers, experimental data were taken from the Livermore, available in EXFOR database^[28]

Download full size

View in Article

Fig. 5. Cross-sections prediction of $(γ, n)$ and $(γ, 2 n)$ reaction channel of ${}^{175} L u$ by Bayesian neural networks with different hidden layers, experimental data were taken from the Saclay, available in EXFOR database^[28]

Download full size

View in Article

Table 1. Training set and verification set, all experimental data were taken from the Livermore or Saclay, available in EXFOR database^[36]

View table

View in Article

Table 1. Training set and verification set, all experimental data were taken from the Livermore or Saclay, available in EXFOR database^[36]

实验室Laboratory	训练集Training set	验证集Verification set
Livermore	⁵¹V，⁵⁹Co，⁵⁵Mn，^58，60Ni，^63，65Cu，⁷⁵As，⁸⁹Y，^{90，91，92，94}Zr，¹⁰⁷Ag，¹¹⁵In，^{116，117，118，119，120，124}Sn，¹²⁷I，¹³³Cs，¹³⁸Ba，¹⁴¹Pr，¹⁵³Eu，¹⁵⁹Tb，¹⁶⁰Gd，¹⁶⁵Ho，¹⁸¹Ta，¹⁸⁶W，^{186，188，189，190，192}Os，^{206，207，208}Pb，²⁰⁹Bi，²³²Th，^235，238U，²³⁹Pu	¹⁹⁷Au
Saclay		¹⁷⁵Lu

Table 2. The time required to train the model 100 000 times for different hidden layers
View table
View in Article
Table 2. The time required to train the model 100 000 times for different hidden layers
神经元×层数Neural number×Layer number 训练次数Training number 训练时间Training time / s
300×2 100 000 234
300×3 289
300×4 318

Tools

Get Citation

Copy Citation Text

Qiankun SUN, Yue ZHANG, Zirui HAO, Hongwei WANG, Gongtao FAN, Hanghua XU, Longxiang LIU, Kaijie CHEN, Sheng JIN, Zhenwei WANG, Mengke XU, Xiangfei WANG. Study of photoneutron reaction based on physics-informed Bayesian neural network[J]. NUCLEAR TECHNIQUES, 2025, 48(5): 050001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Special Topics on Applications of Machine Learning in Nuclear Physics and Nuclear Data

Received: Dec. 6, 2024

Accepted: --

Published Online: Jun. 26, 2025

The Author Email: Hongwei WANG (王宏伟)

DOI:10.11889/j.0253-3219.2025.hjs.48.240501

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Training set and verification set, all experimental data were taken from the Livermore or Saclay, available in EXFOR database[36]

Table 1. Training set and verification set, all experimental data were taken from the Livermore or Saclay, available in EXFOR database[36]

Table 2. The time required to train the model 100 000 times for different hidden layers

Table 2. The time required to train the model 100 000 times for different hidden layers

Table 1. Training set and verification set, all experimental data were taken from the Livermore or Saclay, available in EXFOR database^[36]

Table 1. Training set and verification set, all experimental data were taken from the Livermore or Saclay, available in EXFOR database^[36]