[10a-N304-1~10] 18.4 Methods, Theory, Foundations, Others

9:00 AM - 9:15 AM JST(12:00 AM - 12:15 AM UTC)

[10a-N304-1] Probabilistic Network Estimation of Feature Dependency Structures under Uncertainty

〇Noriaki Ozaki¹ (1.Murata Mfg.)

Keywords:

Dependency network,Exogeneity and dependence-strength evaluation,Probabilistic sampling

Understanding data structure—specifically, relationships among features—is important in engineering. However, XAI primarily focuses on relationships between the target variable and features, and causal analysis requires strong assumptions, making them difficult to apply in practice. Therefore, in this study we propose a method to visualize dependency networks using exogeneity and dependence strength as indicators. The method employs probabilistic sampling to curb overinterpretation and is characterized by presenting only reliable global structures.