On Training Efficiency and Computational Costs of a Feed Forward Neural Network: A Review

Table 5

Summary of fuzzy logic AF.

Ref.	Method	Convergence	Precision	Computational costs	Notes
[27, 31]	Fuzzy-tanh	20 Epochs (up to 4x faster than tanh on the same problem)	MAE = 0.039 (2.5x more accurate than tanh); 93% accuracy (classification problem, 1% more than a classic MLP)	N/A
[28]	Type 2 Fuzzy	41 Epochs (5x faster than tanh on the same problem)	MAE = 0.35	N/A	Backpropagation with learning rate α = 0.25
[32]	Fuzzy-tanh 2	N/A	RMSE = 0.0116 (comparable to tanhon the same problem); 95–98% accuracy (classification problem, comparable to tanh on the same problem)	N/A	Trained with extreme machine learning algorithm