Non-Uniform distribution for colsample

Hello,
I would like to propose a simple modification that can bring good improvement to the algorithm. Select features using a non-uniform distribution for each decision tree. For example, suppose there are 5 features and a uniform distribution, the probability of each feature to be selected is 0.2. But let suppose I want the first feature to be selected 60% of the time. So the non-uniform distrib. could be [0.6, 0.1, 0.1, 0.1, 0.1].
The parameter could be the probability (ps) of each feature to be selected at each iteration (sum(ps)=1). If user don’t send the parameter the algorithm choose it uniform by default (ps = 1/#features).
Giba