Table 1.

Summary of DL models and formulas

Model	Explanation	Formula	Denotation
DNNs	Feed-forward neural networks with multiple hidden layers and activation functions. Approximate nonlinear transformations for specific goals.		: output of neuron , : output of neuron , : weight of neuron , : bias,: activation function
AEs	Deep generative models for dimensionality reduction. Encode input data into latent variables and reconstruct input data.		: latent variable, : encoder network, : reconstructed input, : decoder network
VAEs	Encode inputs as distributions over the latent space. Learn latent features through multi-layer neural networks.		: latent variable, : conditional distribution, : reconstructed input
CNNs	Supervised models for image processing. Extract features from multidimensional input data using convolutional and pooling layers.		: feature map, : input matrix, : kernel
GNNs	Generalized models for graph data processing. Aggregate and transform node information through various network architectures like graph convolution network (GCN) and graph attention network (GAT).		: hidden layer, : layer, : adjacency matrix, : function that aggregates and transforms the hidden state of nodes.