Summary of Conlux: Concept-based Local Unified Explanations, by Junhao Liu et al.
ConLUX: Concept-Based Local Unified Explanations
by Junhao Liu, Haonan Yu, Xin Zhang
First submitted to arxiv on: 16 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 This paper proposes a novel framework called that provides unified concept-based local explanations for any machine learning model. The authors highlight the limitations of existing model-agnostic explanation techniques, which often generate unfaithful and difficult-to-understand explanations. To address this issue, they develop a method to automatically extract high-level concepts from large pre-trained models, allowing them to extend existing local model-agnostic techniques to provide concept-based explanations. The authors demonstrate the effectiveness of by applying it to four different explanation techniques (LIME, Kernel SHAP, Anchor, and LORE) on text and image models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a special tool that helps us understand how machine learning models work. Right now, we have tools that can explain simple features like words or pixels, but these explanations don’t really match what the model is thinking. This new tool takes the big ideas from pre-trained models and uses them to make better explanations for any type of model. The authors test this tool with four different ways of explaining things and show that it works well on text and image recognition tasks. |
Keywords
* Artificial intelligence * Machine learning
