C2_W4_Assignment_build_tree_recursive

Course Q&A Machine Learning Specialization Advanced Learning Algorithms
1

I was just curious if it is better to build the decision tree using while loop since we are using the term recursively to make a more generalized function rather than calling it three times in the implementation.

I appreciate if some can help me out with my pseudo-code:

def build_dt(X, y, root_node_indices , max_depth=2, ):

    # take care of the root_node outside while loop (initialization step) (depth = 0 perhaps???)
    best_feature_root = get_best_split(X, y, root_node_indices)
    left_indices, right_indices = split_dataset(X, root_node_indices, best_feature_root)
    X_left, y_left = X[:, left_indices], y[left_indices]
    X_right, y_right = X[:, right_indices], y[right_indices]
    current_depth = 0

    while current_depth <= max_depth:
        
        best_feature_left = get_best_split(X_left, y_left, left_indices)
        best_feature_right = get_best_split(X_right, y_right, left_right)

        
        left_indices, _ = split_dataset(X_left, left_indices, best_feature_left)
        _, right_indices = split_dataset(X_right, right_indices, best_feature_right)

        X_left, y_left = X[:, left_indices], y[left_indices]
        X_right, y_right = X[:, right_indices], y[right_indices]
        current_depth =+ 1

Is this a correct implementation, or should we logically also fetch the root_node inside the while loop if we consider max_depth=2.

Cheers,

2

Hello @mrgransky,

Let’s just focus on this first:

In the code below, how many times do you think do_print will be called? You can verify your answer by running it. It is definitely not two times :wink:

def do_print(x):
    print(x)
    x += 1
    if x < 10:
        do_print(x)

do_print(0)

What controls the number of times do_print is called? In other words, if you want it to print 100 numbers, what would you change?

Raymond