We will now be using the top 2 values of predictions_top5 to produce the confusion matrix. We will create the confusion matrix by considering the correct section jj and the two sections with the two highest scores i1,i2i1,i2 where score(i1)>score(i2)score(i1)>score(i2).

• A prediction is correct if j=i1j=i1. correct examples are ignored.
• The predicion i2i2 is confused with i1i1 if j=i2j=i2
• If ji1ji1 and ji2ji2 then the example is ignored.

Examples

We consider two scenarios where the possible sectors/categories are (0,1,2,3)

Scenario 1

Sample Input y_label = 3 top2_predictions = [2, 3]

Output confusion_matrix: [0 0 0 0] [0 0 0 0] [0 0 0 1] [0 0 0 0] Explanation: The first example has top two predictions: (2, 3) and the label 3. So we add 1 to the position (2,3).

Scenario 2

Sample Input Say we have 7 sample prediction values y_label = [3, 2, 3, 1, 0, 2, 2] top2_predictions = [[2, 3], [0, 3], [2, 3], [1, 2], [3, 0], [2, 0], [3, 2]]

Output confusion_matrix: [0 0 0 0] [0 0 0 0] [0 0 0 2] [1 0 1 0]

Explanation: The second example has 7 sample predictions:

1. In two scenarios 2 is predicted in place of 3 and the cell (2,3) is incremented twice.
2. In two other scenarios we have a case where the second prediction is right. The corresponding cells {(3,0) and (3,2)} are incremented once each.
3. In two scenarios, the first element is the correct prediction. Thus no cell is incremented.
4. In one scenario, neither of the top 2 predictions is correct. Thus no cell is incremented.

Note: y_label is the same as y_valid here.

All but one of the lines of the function have been filled. You need only fill the one missing line that updates the confusion matrix if a confusion (i.e., sector that should be labeled top but is predicted second) happenes during a cycle

def get_confusion_matrix(predictions_top5, y_valid):

 confusion_matrix = np.zeros((11,11), dtype=int64)

 i=0

 for entry in predictions_top5[:, :2]:

 if entry[1] == y_valid[i]:

 # update confusion_matrix

 ###

 ### YOUR CODE HERE

 ###

 i += 1

 return confusion_matrix