From Pixels to Perception: Interpretable Predictions via Instance-Wise Grouped Feature Selection

Abstract

Understanding the decision-making process of machine learning models provides valuable insights into the task, the data, and the reasons behind a model’s failures. In this work, we propose a method that performs inherently interpretable predictions through the instance-wise sparsification of input images. To align the sparsification with human perception, we learn the masking in the space of semantically meaningful pixel regions rather than on pixel-level. Additionally, we introduce an explicit way to dynamically determine the required level of sparsity for each instance. We show empirically on semi-synthetic and natural image datasets that our inherently interpretable classifier produces more meaningful, human-understandable predictions than state-of-the-art benchmarks.

Cite

Text

Vandenhirtz and Vogt. "From Pixels to Perception: Interpretable Predictions via Instance-Wise Grouped Feature Selection." Proceedings of the 42nd International Conference on Machine Learning, 2025.

Markdown

[Vandenhirtz and Vogt. "From Pixels to Perception: Interpretable Predictions via Instance-Wise Grouped Feature Selection." Proceedings of the 42nd International Conference on Machine Learning, 2025.](https://mlanthology.org/icml/2025/vandenhirtz2025icml-pixels/)

BibTeX

@inproceedings{vandenhirtz2025icml-pixels,
  title     = {{From Pixels to Perception: Interpretable Predictions via Instance-Wise Grouped Feature Selection}},
  author    = {Vandenhirtz, Moritz and Vogt, Julia E},
  booktitle = {Proceedings of the 42nd International Conference on Machine Learning},
  year      = {2025},
  pages     = {60833-60856},
  volume    = {267},
  url       = {https://mlanthology.org/icml/2025/vandenhirtz2025icml-pixels/}
}