Complexity-Penalized PyTorch Loss | Dataford Interview Questions

Implement a custom loss function in PyTorch that penalizes model complexity as well as prediction error.

Return the scalar mean squared error plus complexity_weight times the sum of squared values across all parameter tensors. Implement the equivalent calculation using nested Python lists so JSON test inputs can be evaluated; the function signature is custom_loss(predictions, targets, parameters, complexity_weight). For example, predictions [1, 2], targets [0, 4], parameters [[1, -2], [0.5]], and weight 0.1 produce 3.025. Inputs contain matching prediction and target shapes, and the complexity weight is nonnegative.

Constraints

predictions and targets contain the same number of numeric values
predictions contains at least one numeric value
parameters is a list of numeric values or nested lists
complexity_weight >= 0

Function Signature

def custom_loss(predictions, targets, parameters, complexity_weight):

Your question is Complexity-Penalized PyTorch Loss. 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

Implement a custom loss function in PyTorch that penalizes model complexity as well as prediction error.

Constraints

predictions and targets contain the same number of numeric values
predictions contains at least one numeric value
parameters is a list of numeric values or nested lists
complexity_weight >= 0

Function Signature

def custom_loss(predictions, targets, parameters, complexity_weight):

Your question is Complexity-Penalized PyTorch Loss. 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.