Custom Python Loss Function | Dataford Interview Questions

Write a function to implement a custom loss function in Python using standard libraries.

Implement the mean Huber loss for equal-length numeric sequences and return both the scalar loss and its gradient with respect to each prediction. For residual r = prediction - target, use 0.5 * r² when abs(r) <= delta, otherwise delta * (abs(r) - 0.5 * delta).

Contract

custom_loss(y_true, y_pred, delta) returns (mean_loss, gradients).

Constraints

1 <= len(y_true) = len(y_pred) <= 1000
y_true and y_pred contain integers or floating-point values
delta > 0
The input sequences are nonempty and have equal length

Function Signature

def custom_loss(y_true, y_pred, delta):

Your question is Custom Python Loss Function. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem