Implement Dice Loss

Dice loss comes from image segmentation, where most pixels are background and a plain per-pixel loss barely notices the small foreground region. It instead measures the overlap between the predicted and true masks directly.

\text{Dice} = \frac{2 \sum (y_{\text{true}} \cdot y_{\text{pred}}) + \text{smooth}}{\sum y_{\text{true}} + \sum y_{\text{pred}} + \text{smooth}}

Task: write dice_loss(y_true, y_pred, smooth) returning 1 - Dice, rounded to 4 decimal places.

y_true and y_pred are flat lists of values in [0, 1] — a mask and a prediction, one value per pixel.
smooth keeps the loss defined even when both masks are entirely empty; without it that case would divide by zero.

Implement Dice Loss

Examples

Hints