**arXiv ID:** 2305.00543 **Authors:** Geetanjali Bihani, Julia Taylor Rayz **Published:** 2023-04-30T18:06:14Z **Abstract:** Neural network-based decisions tend to be overconfident, where their raw outcome probabilities do not align with the true decision probabilities. Calibration of neural networks is an essential step towards more reliable deep learning frameworks. Prior metrics of calibration error primarily utilize crisp bin membership-based measures. This exacerbates skew in model proba...
Installs into .claude/skills of the current project.
Sponsors
SSkills DirectorySkills Directory
Know which skills are safe — weekly.
SSkills DirectorySkills Directory
Is that skill safe to install?
Are you the author of Calibration Error Estimation Using Fuzzy Binning?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/hiyenwong-calibration-error-estimation-using-fuzzy-binning)
# Calibration Error Estimation Using Fuzzy Binning
**arXiv ID:** 2305.00543
**Authors:** Geetanjali Bihani, Julia Taylor Rayz
**Published:** 2023-04-30T18:06:14Z
**Abstract:**
Neural network-based decisions tend to be overconfident, where their raw outcome probabilities do not align with the true decision probabilities. Calibration of neural networks is an essential step towards more reliable deep learning frameworks. Prior metrics of calibration error primarily utilize crisp bin membership-based measures. This exacerbates skew in model probabilities and portrays an incomplete picture of calibration error. In this work, we propose a Fuzzy Calibration Error metric (FCE) that utilizes a fuzzy binning approach to calculate calibration error. This approach alleviates the impact of probability skew and provides a tighter estimate while measuring calibration error. We compare our metric with ECE across different data populations and class memberships. Our results show that FCE offers better calibration error estimation, especially in multi-class settings, alleviating the effects of skew in model confidence scores on calibration error estimation. We make our code and supplementary materials available at: https://github.com/bihani-g/fce
## Skill Description
This skill is generated from the arXiv paper: Calibration Error Estimation Using Fuzzy Binning (2305.00543).
## How to Use
[To be filled in by the user or by future automation]
## References
- [arXiv:2305.00543](http://arxiv.org/abs/2305.00543v2)
