Asking Numpy. Why i can get result this?

Let's break this down piece by piece.

1. np.unique(y)

   np.unique returns all of the unique elements of the given array. In this case, that's [0, 1] (since there are only two unique elements in y, 0 and 1).

   So running

   for label in np.unique(y): 

   will iterate twice. The first time label will equal 0, and the second time it will equal 1. You can inspect this for yourself by running

   for label in np.unique(y): print(label) 

2. y == label

   If you run a comparison against an array, numpy will return a boolean array of the same size. So on the first iteration of the for loop, it'll be running y == 0, which gives us [ True False True False], since the first and third element in y are 1. Then on the second iteration, it'll be running y == 1, giving us the inverse, [False True False True].

   You can inspect this for yourself by running:

   for label in np.unique(y): print(y == label) 

3. X[y == label]

   Now that we know what the rest of the logic is doing, the last piece is determining what happens when passing the y == label as a "selector" for elements in X. When you pass a boolean array as a selector, you're telling numpy which elements of the original array you want to return. In this case, X is a 2-dimensional array (size 4x4), and if we pass an array of length 4, we're telling numpy which rows we want to select.

   As mentioned above, on the first iteration y == label is [ True False True False], so we're saying we want the first and third rows of X. So print(X[y == label]) gives

   [[0 1 0 1] [0 0 0 1]] 

   And then on the second iteration, per the above, y == label will be selecting the second and fourth rows, giving:

   [[1 0 1 1] [1 0 1 0]] 

Hope that clarifies!