Table A3.

Model‐specific explainable artificial intelligence (XAI) methods.

Model‐specific method	Acronym	Black‐box model	Type of XAI	Scope of XAI	Technique
Class activation map [121]	CAM	Convolutional neural network (CNN)	Feature importance	Local	Propagation‐based
Gradient‐weighted class activation mapping [122]	Grad‐CAM	CNN	Feature importance	Local	Propagation‐based
Gradient‐weighted class activation mapping++ [123]	Grad‐CAM++	CNN	Feature importance	Local	Propagation‐based
Integrated gradients [138]	IG	All DL models	Feature importance	Local	Propagation‐based
Deep Learning Important FeaTures [139]	DeepLIFT	All DL models	Feature importance	Local	Propagation‐based
Layerwise relevance propagation [124]	LRP	All DL models	Feature importance	Local	Propagation‐based
Deep Taylor decomposition [140]	DTD	All DL models	Feature importance	Local	Propagation‐based
Guided backpropagation [141]	GBP	All DL models	Feature importance	Local	Propagation‐based
Activation maximization [142]	N/A	All DL models	Feature importance	Global	Propagation‐based
Testing with concept activation vectors [143]	TCAV	All DL models	Feature importance	Global	Concept‐based
Model explanation for graph neural networks [144]	GraphLIME	GNN	Feature importance	Local	Surrogate model