.png){kind=icon}
Return
Genetic Algorithm to Understand Image Classification
.png){kind=icon}
PRE
AI
DOI：10.1007/978-3-031-97119-8_41.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper introduces Genetic IPHA, a novel method for Explainable Artificial Intelligence (XAI) aimed at identifying the importance of pixels in image classification tasks performed by neural networks. The proposed approach formulates interpretability as an optimization problem, employing a genetic algorithm to maximize a fitness function and generate masks that highlight either important or unimportant pixels for model predictions. Experiments were conducted using the CIFAR-10 dataset to evaluate and compare Genetic IPHA against established interpretability methods. The results demonstrate that Genetic IPHA consistently outperforms other algorithms, achieving superior accuracy in identifying both relevant and non-relevant pixels. This highlights its potential as a robust and effective solution for enhancing model interpretability.
Keywords:
Explainable artificial intelligence
Genetic algorithm
Image classification
Journal
E
IF:
0
Papers:
191
Citations:
0
Organization
View more organizations
Cited Papers
Tell Model Where to Attend: Improving Interpretability of Aspect-Based Sentiment Classification via Small Explanation Annotations.png)
ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
OAAI
ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)IF0
OAAIShareSaveExplaining Deep Neural Networks and Beyond: A Review of Methods and ApplicationsPROCEEDINGS OF THE IEEE
OAAI
PROCEEDINGS OF THE IEEEIF25.9
OAAIShareSaveShareSaveShareSaveRobust explainability: a tutorial on gradient-based attribution methods for deep neural networks
IF0
PREAI
ShareSaveShareSave View more