TABLE 1.

A summary of the neural networks.

Neural networks	Advantages	Disadvantages	Biomedical tasks
Fully connected neural network	It is widely used at the end of the other neural network models to integrate features and make predictions	It is not easy to process high-dimensional data	Combined with other neural networks, it is widely used in many fields
Convolutional neural network	It can extract highly abstract and complex features from images	It has too many parameters, and the training speed is slow	It is suitable for processing imaging-related tasks, such as clinical imaging
Recurrent neural network	It has a memory function and can effectively process data about sequence and time	Training procedure is difficult and computationally intensive	It is suitable for processing sequence related biomedical data, such as DNA sequence, protein sequence, electronic health records
Autoencoder	It can perform unsupervised learning without using labeled data	It needs a pretraining phase	It is suitable for feature dimensionality reduction or learning effective features from data, such as clinical imaging and genomics
Deep belief network	It can be used for both supervised learning and unsupervised learning	The training process is computationally intensive	It is suitable for automatic feature extraction tasks, such as genomics and drug development