plete. For instance, to provide accurate discrimination between

nd malignant tumours, researchers have employed deep learning

ms to construct predictive models based on either biomarkers or

Coudray , et al., 2018; Munir, et al., 2019; Shen, et al., 2019;

t al., 2020; Pathak, et al., 2020; Kumar and Bakariya, 2021;

al., 2021; Masoudi, et al., 2021].

rent deep learning algorithms have also been considered for

data sets and requirements. The algorithms include different

strategies such as static models like the convolutional neural

and the generative adversarial models as well as the dynamic

ike restricted Boltzmann's machine and the recurrent neural

[Munir, et al., 2019]. Based on a deep neural network model

ed on the image data, the detection accuracy for breast cancer

al., 2019; Lotter, et al., 2021], lung cancer [Coudray , et al., 2018;

nd Bakariya, 2021] and prostate cancer [Masoudi, et al., 2021] as

ome other cancers has been significantly improved. Deep learning

been used to detect differentially expressed genes from a single-

-sequencing count data set [Cui and Wang, 2021], to discriminate

S-CoV-2 genome from other viral genomes [Lopez-Rincon, et al.,

ver, no matter how successful neural network and deep learning

ms are in different areas, the advantages of the linear models have

ome the disadvantages of neural network and deep learning

Heaven, 2019; Waldrop, 2019]. After a neural network model or

ural network model has been well trained, its tolerance capability

doubt. This is not a surprise because most of these models are

ned in house using well-prepared data [Heaven, 2019]. Heaven’s

s that such a model is too complicated so that it may pick up too

ails, making it fragile. George Hinton, the pioneer of the theory

rithms of neural networks and deep neural networks, also had a

“what is missing?” Waldrop has suggested to consider the

ent of separate networks rather than a single network which has

mplicated structure. Even so, deep learning is still making a huge

in many areas including biological/medical pattern analysis. As