Just because you can analyse it doesn't mean that it is better. Deep learning theory is unbelievably garbage compared to the empirical results.

In particular, please show me a worked example of a decision tree meta learning. Because its trivial to show this for DNNs.