.png){kind=icon}
.png){kind=icon}
Return
The Explainability-Performance Coefficient: A New Metric for Model Transparency
.png){kind=icon}
PRE
AI
DOI：10.1007/978-3-032-04558-4_25.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Deep Learning models have shown remarkable performance across multiple domains, yet their lack of interpretability remains a significant challenge. In this work, we propose the Explainability-Performance Coefficient (EP C), a novel metric that quantifies the tradeoff between identifying the most influential input features and preserving model performance, where a higher EP C implies an improved balance between these two critical aspects. Our results show that model-specific explainability techniques yield higher EP C values than other methods such as feature selection. Furthermore, when combined with model-based approaches, regularization significantly improves explainability by effectively reducing the number of relevant features without compromising performance.
Keywords:
Explainability
Explainability of Deep Learning
XAI
Journal
A
IF:
0
Papers:
53
Citations:
0