Concepts de base
DCNFIS combines fuzzy logic and deep learning for improved transparency without sacrificing accuracy.
Résumé
The article introduces DCNFIS, a novel deep network that enhances transparency in AI systems by combining fuzzy logic and deep learning. It outperforms existing CNNs on benchmark datasets and offers explanations through saliency maps derived from fuzzy rules. The paper details the architecture, methodology, evaluation on benchmark datasets, and interpretability analysis using Fashion-MNIST as a case study.
Introduction to DCNFIS combining fuzzy logic and deep learning for enhanced transparency.
Design of DCNFIS architecture with end-to-end training capabilities.
Evaluation of DCNFIS performance on benchmark datasets like MNIST, Fashion-MNIST, CIFAR-10, CIFAR-100.
Interpretability analysis using saliency maps derived from fuzzy rules on Fashion-MNIST dataset.
Stats
DCNFISは既存のCNNを上回る性能を示しました。
DCNFISはMNIST、Fashion-MNIST、CIFAR-10、CIFAR-100のベンチマークデータセットで評価されました。