Summary of Toward Understanding the Disagreement Problem in Neural Network Feature Attribution, by Niklas Koenen and Marvin N. Wright
Toward Understanding the Disagreement Problem in Neural Network Feature Attribution
by Niklas Koenen, Marvin N. Wright
First submitted to arxiv on: 17 Apr 2024
Categories
- Main: Machine Learning (stat.ML)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed paper investigates the inner workings of neural networks by examining feature attribution methods, which assign relevance scores to input variables for model predictions. The study aims to resolve the ongoing debate about which method to use by exploring the fundamental and distributional behavior of explanations. The research also assesses the impact of scaling and encoding techniques on explanation quality and evaluates the efficacy across different effect sizes. Additionally, the paper highlights inconsistencies in rank-based evaluation metrics. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The proposed paper tries to figure out how neural networks work by looking at how they assign importance to each piece of data that goes into a prediction. The study wants to settle the argument about which way is best to do this by studying what makes explanations good or bad. It also looks at how different ways of preparing and scaling the data affect how well it explains things, and whether certain methods are better than others. |
