RLT.default<- function(formula, traindata, model, print_summary = FALSE, use_cores = 1, 
				ntrees=100, mtry = ncol(traindata)/3, nmin = 6, resample = TRUE, resample_prob = 0.75,
				split_gen = "Rank", nspliteach = 10, select_method = 1, 
				reinforcement = FALSE, muting = -1, muting_percent = 0.2, protectVar = sqrt(ncol(traindata)), combsplit_th = 0.5, combsplit = 1)



model : please specify "regression" or "classification"; "survival" is under test.

use_cores : number of CPU threads can be used

ntrees : number of trees

mtry : number of variable sampled at each node; for embedded model, its the same value

nmin : minimum number of observations to produce a split

resample : randomly sample from the training dataset to fit the model, for extreamly randomized trees, set to "FALSE". However, you would not be able to compute variable importance.

resample_prob : proportion of observations sampled to fit each tree (without replacement).

split_gen : how to splitting point is generated: "Uniform", uniformly from (min, max) for a variable; "Rank", random rank; "Best", search for all possible splits.

nspliteach : if use random split generation, i.e. "Uniform" or "Rank", how many number of splits for each variable are genarated and compared.

select_method : 1 for regression, 2 for classification. you dont have to specify this when "model" is specified.


########### reinforcement learning features: use when you truly understand what you want to do... 


reinforcement : set to TRUE if you want to do reinforcement learning at each split

muting : muting mechanicsm: 0 if no muting; -1 mute by proportion of number of variables; if >1, mute a fixed number of variables at each split.

muting_percent : if "muting" is set to -1, mute this percent of variables at each split. The defaul it 0.2 (20%).

protectVar : number of protected variables. 

combsplit : integers > 1 for the number of variables used in linear combination split (default is 1; can not be larger than 5 at the current stage of this package).

combsplit_th : range 0-1. threshold for which variable can be used in the linear combination split (as the proportion of largest variable importance; suggest value 0.5)