omputational algorithms for pattern analysis [Rumelhart, et al.,

shop, 1996]. MLP rooted from the human brain research and the

computational technology. It is also called an information

g system because it can be used for knowledge discovery and

association through a large-scale data mining process. The most

ase is that MLP can be used to discover the pattern underlining

to re-construct it without a priori knowledge about what the

unction looks like. The most important feature that makes MLP

ent intelligent information processing system is its power of

nowledge or intelligence learned from a data set. Due to this

ness, MLP has been widely used for pattern discovery and

on in many areas including biology and medicine. For instance,

n used for amelioration of inflammatory bowel disease based on

del [Sabater, et al., 2019]. Another recent study using MLP has

investigation of generating single-cell RNA-seq through learning

gene-gene relationship of complex, multiple cell type samples

et al., 2020].

he structure of MLP

21 shows one example of MLP structure, in which there are three

computing units called neurons. They are a layer of the input

ݔଵand ݔଶ), a layer of the hidden neurons (ݖଵand ݖଶ) and a layer

tput neuron (y). Suppose

ݖଵൌ݂ଵሺݔଵ, ݔଶሻ

ݖଶൌ݂ଶሺݔଵ, ݔଶሻ

ݕൌ݂ଷሺݖଵ, ݖଶሻ

(3.29)

ependent variable y is then a complex function of the independent

ݔଵ and ݔଶ,

ݕൌ݂ଷሼ݂ଵሺݔଵ, ݔଶሻ, ݂ଵሺݔଵ, ݔଶሻሽ

(3.30)