Summary of Interpretability in Symbolic Regression: a Benchmark Of Explanatory Methods Using the Feynman Data Set, by Guilherme Seidyo Imai Aldeia and Fabricio Olivetti De Franca (federal University Of Abc)
Interpretability in Symbolic Regression: a benchmark of Explanatory Methods using the Feynman data set
by Guilherme Seidyo Imai Aldeia, Fabricio Olivetti de Franca
First submitted to arxiv on: 8 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 aims to develop a benchmark scheme for evaluating explanatory methods in regression tasks, particularly symbolic regression models. The authors highlight the importance of interpretability in machine learning models, citing its role in verifying model properties and improving fairness. They argue that existing model-agnostic explanatory methods can provide explanations for black-box models, but there is a need for a rigorous evaluation and comparison of these methods’ quality. To address this gap, the paper proposes a benchmark scheme that combines symbolic regression models with popular explanation methods, evaluating their performance using various measures. The authors perform experiments on 100 physics equations, demonstrating the potential of symbolic regression models as an alternative to white-box and black-box approaches. Their results show that Partial Effects and SHAP are robust explanation models, while Integrated Gradients can be unstable when used with tree-based models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about a new way to evaluate how well machine learning models work. Right now, it’s hard to know why some models make certain predictions. This paper suggests creating a special test set that shows how well different methods can explain what the model did. They tested this idea on 100 math problems and found that one type of model, called symbolic regression, was very good at both making accurate predictions and explaining its answers. The best explanation methods were Partial Effects and SHAP. |
Keywords
* Artificial intelligence * Machine learning * Regression
