Brier score
Beginner Netflix, Spotify+5
The problem, from first principles
Log loss is sharp about confident mistakes, and sometimes that sharpness is more than you want. The Brier score is the mean squared error between the predicted probability and the 0/1 outcome. It is bounded between 0 and 1, so a single extreme miss cannot dominate it the way it can in log loss.
Implement brier_score(y_true, y_prob) for binary labels, where y_prob is the predicted probability of class 1.
Return mean((p − y)²).
Constraints
y_truevalues must be 0 or 1, andy_probvalues must lie in[0, 1]. Otherwise raiseValueError.- Inputs must have the same length and be non-empty.
Hints
Open one at a time. Each gives away a little more than the last.
Hint 1
Subtract the labels from the probabilities element by element, then square.
Hint 2
np.mean of the squared differences is the whole answer.
Keywords
classical-mlevaluationclassificationcalibration
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.