I have made changes according to the mistakes you pointed out. And answers for the questions:
- In the algorithm shown, how is pick happening? is it random?
-- Yes, here the algorithm randomly picks a condition on which it will split the input data samples.
- "Thus, to make a decision tree one can search the entire space for the smallest decision tree." Where do you get the search space from? must you create all the trees first?
-- Yes, the naive way is to search the entire decision tree space by making them and then pick the smallest one. But as the state space grows exponentially, we greedily search while minimizing the error. The description of this greedy search algorithm is provided in my Wiki page.
Thanks, Ekta Aggarwal