Implement a Custom Loss Function

medium (<30 mins) pytorch loss functions regression implementation

this month by E

Create a custom loss function called MeanAbsolutePercentageError (MAPE) in PyTorch. It should:

Take predictions and targets as input tensors.
Compute $\frac{1}{n} \sum_{i} \frac{| y_{i} - {\hat{y}}_{i} |}{| y_{i} | + ϵ}$ , with $ϵ = 1 e - 8$ to avoid division by zero.
Return a scalar tensor.

Test it on a random regression output.