Unit 2: Deep feedforward networks

Question 1

Accuracy

Answer 1

The proportion of correct predictions

Question 2

Error rate

Answer 2

The preportion of incorrect predictions

Question 3

TPR

Answer 3

The True positive rate is the proportion of actual positives that are correctly predicted.

AKA Sensitivity

Question 4

False positive rate

Answer 4

The proportion of actual negatives that are predicted as positive.

Question 5

Receiver Operating Characteristic (ROC) curve

Answer 5

The True Positive Rate, against the False Positive rate, as a curve.

Question 6

Recall

Answer 6

The actual positives classified correctly.

recall = TP / (TP + FN)

Question 7

Precision

Answer 7

The proportion of predicted positives classified correctly

precision = TP / (TP + FP)

Question 8

F-score

Answer 8

The harmonic mean of precision and recall.

F = 2 x (precision x recall) / (precision + recall)

Question 9

Why use a harmonic mean for the F-score?

Answer 9

To penalise situations where either the precision or recall are low.

Question 10

Balanced accuracy

Answer 10

The mean of the recall values for each class.