Summary of Improving Explainability Of Softmax Classifiers Using a Prototype-based Joint Embedding Method, by Hilarie Sit et al.
Improving Explainability of Softmax Classifiers Using a Prototype-Based Joint Embedding Method
by Hilarie Sit, Brendan Keith, Karianne Bergen
First submitted to arxiv on: 2 Jul 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 A prototype-based approach to improve explainability of softmax classifiers is proposed, providing understandable prediction confidence generated by stochastic sampling of prototypes. This demonstrates potential for out-of-distribution detection (OOD). By modifying the model architecture and training, predictions are made using similarities to any set of class examples from the training dataset. This enables sampling for prototypical examples that contributed to the prediction, providing an instance-based explanation for the model’s decision. Additionally, learning relationships between images in the training dataset through relative distances within the model’s latent space yields a metric for uncertainty better suited for detecting out-of-distribution data than softmax confidence. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper proposes a new way to explain why a model made a certain prediction using prototypes from the training data. The approach helps identify which specific examples from the training set contributed to the prediction, making it easier to understand how the model works. This could be useful for detecting when the model is working outside of its usual range and can’t make accurate predictions. |
Keywords
* Artificial intelligence * Latent space * Softmax
