Structure learning of Bayesian network using coordinate-descent
algorithm. This algorithm is designed for discrete network assuming a multinomial data set,
and we use a multi-logit model to do the regression.
The algorithm is described in Gu, Fu and Zhou (2016)

An algorithm to learn structure of discrete Bayesian network, this package can deal with observational data, interventional data, or a misture of both.

`cd.run`

is the main function to run coordinate descent algorithm. With the`adaptive`

option, users may choose to use regular group lasso penalty, or adaptive group lasso penalty.`max_lambda`

is a function to calculate the maximum value of lambda that will penalized all edges to zero.