ChaosMining: A Benchmark to Evaluate Post-Hoc Local Attribution Methods in Low SNR Environments

TL;DR

This paper introduces ChaosMining, a benchmark for evaluating post-hoc local attribution methods in low SNR environments, revealing their strengths and limitations across different models and data types.

Contribution

It provides a comprehensive benchmark dataset and analysis framework for attribution methods in noisy, real-world scenarios, and proposes an improved recursive feature elimination algorithm for neural networks.

Findings

Attribution methods struggle with low SNR data.

The new RFE extension improves feature selection accuracy.

Scalability remains a challenge for the proposed method.

Abstract

In this study, we examine the efficacy of post-hoc local attribution methods in identifying features with predictive power from irrelevant ones in domains characterized by a low signal-to-noise ratio (SNR), a common scenario in real-world machine learning applications. We developed synthetic datasets encompassing symbolic functional, image, and audio data, incorporating a benchmark on the {\it (Model \(\times\) Attribution\(\times\) Noise Condition)} triplet. By rigorously testing various classic models trained from scratch, we gained valuable insights into the performance of these attribution methods in multiple conditions. Based on these findings, we introduce a novel extension to the notable recursive feature elimination (RFE) algorithm, enhancing its applicability for neural networks. Our experiments highlight its strengths in prediction and feature selection, alongside limitations in scalability. Further details and additional minor findings are included in the appendix, with extensive discussions. The codes and resources are available at \href{https://github.com/geshijoker/ChaosMining/}{URL}.