解释用于马射线检测和识别的机器学习模型
Mark S Bandstra1, Joseph C Curtis1, James M Ghawaly2
1Nuclear Science Division, Lawrence Berkeley National Laboratory, Berkeley, California, United States of America.
PloS one
|June 20, 2023
概括
像SHAP这样的可解释AI方法可以提高对复杂的马射线光谱分析模型的理解. 新技术提高了放射学数据应用的模型解释性和准确性.
科学领域:
- 核物理 核物理 核物理
- 数据科学数据科学数据科学
- 人工智能的人工智能
背景情况:
- 复杂的预测模型越来越多地用于马射线光谱分析.
- 可解释的人工智能 (XAI) 技术正在出现,以理解这些模型.
- 正在探索基于梯度 (例如,Grad-CAM) 和黑子 (例如,LIME,SHAP) 的方法.
研究的目的:
- 将XAI方法与玛射线光谱数据进行比较和调整.
- 为了评估不同XAI技术的准确性.
- 提出新的方法来产生反事实解释.
主要方法:
- 使用一个神经网络模型,在合成的NaI(Tl) 城市搜索数据上进行训练.
- 基于梯度和黑盒XAI方法的比较.
- 开发了一种使用直角投影进行反事实解释的技术.
主要成果:
- 黑子方法,LIME和SHAP,在解释模型预测方面表现出高准确度.
- 由于其最小的超参数调要求,建议使用SHAP.
- 成功演示了一种用于生成反事实解释的新技术.
结论:
- XAI方法可以有效地适应用于马射线光谱分析.
- 在这个领域,SHAP为模型可解释性提供了一个强大而准确的解决方案.
- 提出的反事实解释技术增强了对模型行为的理解.
相关概念视频
01:20Maxwell-Boltzmann Distribution: Problem Solving
1.6K
Individual molecules in a gas move in random directions, but a gas containing numerous molecules has a predictable distribution of molecular speeds, which is known as the Maxwell-Boltzmann distribution, f(v).
This distribution function f(v) is defined by saying that the expected number N (v1,v2) of particles with speeds between v1 and v2 is given by
This distribution function f(v) is defined by saying that the expected number N (v1,v2) of particles with speeds between v1 and v2 is given by
1.6K
01:19Gas Chromatography: Types of Detectors-II
436
In gas chromatography, different detectors are employed to meet specific analytical needs. These detectors are often categorized based on their detection mechanisms and the types of compounds they are best suited to analyze. Thermal Conductivity Detectors (TCD), Flame Ionization Detectors (FID), and Electron Capture Detectors (ECD) represent common categories, each with unique operating principles and applications. However, beyond these, several other detectors are designed for more specialized...
436