paper-with-me

홈 › Papers

A Guide to Feature Importance Methods for Scientific Inference

2024-04-19 · Fiona Katharina Ewald, Ludwig Bothmann, Marvin N. Wright, Bernd Bischl, Giuseppe Casalicchio, Gunnar König

While machine learning (ML) models are increasingly used due to their high predictive power, their use in understanding the data-generating process (DGP) is limited. Understanding the DGP requires insights into feature-target associations, which many ML models cannot directly provide due to their opaque internal mechanisms. Feature importance (FI) methods provide useful insights into the DGP under certain conditions. Since the results of different FI methods have different interpretations, selecting the correct FI method for a concrete use case is crucial and still requires expert knowledge. This paper serves as a comprehensive guide to help understand the different interpretations of global FI methods. Through an extensive review of FI methods and providing new proofs regarding their interpretation, we facilitate a thorough understanding of these methods and formulate concrete recommendations for scientific inference. We conclude by discussing options for FI uncertainty estimation and point to directions for future research aiming at full statistical inference from black-box ML models.

📄 PDF Abstract BibTeX arXiv:2404.12862

Code (1)

slds-lmu/paper_2024_guide_fi 공식 구현

Tasks

Feature Importance

Similar Papers 제목 키워드 기반

IDLat: An Importance-Driven Latent Generation Method for Scientific Data

2022-08-05 · Jingyi Shen, Haoyu Li, Jiayi Xu, Ayan Biswas 외

Deep learning based latent representations have been widely used for numerous scientific visualization applications such as isosurface similarity analysis, volume rendering, flow field synthesis, and data reduction, just…

Data Visualization

Physics guided machine learning using simplified theories

2020-12-18 · Suraj Pawar, Omer San, Burak Aksoylu, Adil Rasheed 외

Recent applications of machine learning, in particular deep learning, motivate the need to address the generalizability of the statistical inference approaches in physical sciences. In this letter, we introduce a modular…

BIG-bench Machine Learning

Doctor Rashomon and the UNIVERSE of Madness: Variable Importance with Unobserved Confounding and the Rashomon Effect

2025-10-14 · Jon Donnelly, Srikar Katta, Emanuele Borgonovo, Cynthia Rudin arxiv

Variable importance (VI) methods are often used for hypothesis generation, feature selection, and scientific validation. In the standard VI pipeline, an analyst estimates VI for a single predictive model with only the ob…

Model-Agnostic Confidence Intervals for Feature Importance: A Fast and Powerful Approach Using Minipatch Ensembles

2022-06-05 · Luqin Gan, Lili Zheng, Genevera I. Allen

To promote new scientific discoveries from complex data sets, feature importance inference has been a long-standing statistical problem. Instead of testing for parameters that are only interpretable for specific models, …

BIG-bench Machine LearningEnsemble LearningFeature ImportanceInterpretable Machine Learning

Neural Importance Sampling for Rapid and Reliable Gravitational-Wave Inference

2022-10-11 · Maximilian Dax, Stephen R. Green, Jonathan Gair, Michael Pürrer 외

We combine amortized neural posterior estimation with importance sampling for fast and accurate gravitational-wave inference. We first generate a rapid proposal for the Bayesian posterior using neural networks, and then …

Diagnostic