Binary Cross Entropy

Binary cross entropy (BCE; aka log-loss) is a loss function associated with binary classification problems. The formula for BCE is:

$$\text{loss}=\frac{1}{N}\sum _{i=1}^N{y}_i\cdot \log (\widehat{y_i})+(1-y_i)\cdot \log (1-\widehat{y_i})$$

where $\widehat{y_i}$ is the predicted probability that $y_i=1$ and $y_i$ is either 0 or 1.