Problems

1 / 1

Implement Micro-F1

Metrics & Evaluation

Easy

Compute the micro-averaged F1 score for equal-length sequences of single-label multiclass predictions. First aggregate true positives, false positives, and false negatives across every class, then compute

F_{1,\mathrm{micro}} = \frac{2TP}{2TP + FP + FN}

Here, $TP$ , $FP$ , and $FN$ are the totals across all classes. Return the score as a Python float rounded to four decimal places.

Loading visualization...

Examples

Input: y_true = [0, 1, 1], y_pred = [0, 1, 0]

Output: 0.6667

Explanation: Across all classes there are two true positives, one false positive, and one false negative.

Input: y_true = [0, 1, 2, 2], y_pred = [0, 1, 2, 2]

Output: 1.0

Input: y_true = [2, 2, 1, 0], y_pred = [1, 2, 1, 0]

Output: 0.75

Hint 1

sum(actual == predicted for actual, predicted in zip(y_true, y_pred)) counts correct single-label predictions.

Hint 2

For single-label multiclass data, every mismatch contributes one false positive and one false negative.

Requirements

y_true and y_pred contain one integer class label per sample
Aggregate counts across every class
Return a Python float rounded to four decimal places

Constraints

y_true and y_pred have the same nonzero length
At most $100{,}000$ labels are provided
Do not use an external machine learning library

Try Similar Problems

Classification Metrics Auc Iou Bounding Box Confusion Matrix Norm Log Loss Per Sample

Prev Next