If you're interested, in my thesis I induced l1-regularized decision trees through a boosting style approach. Adding an l1 term and maximizing the gradient led

If you're interested, in my thesis I induced l1-regularized decision trees through a boosting style approach. Adding an l1 term and maximizing the gradient led to sparse tree.