使用 NumPy 写出分段函数？

Question

I'm new to NumPy and trying to figure out how can I write down piecewise-defined function. I'm having a derivative of MSE_Loss function using ReLU as activation function:

I have all parameters I need. My question is: how to use NumPy to handle piecewise functions (preferable without using loops)?

EDITED : For example, I can write down the following function using NumPy following way:

def loss_derivative(X, y, w, b):
    n = len(y)
    sigma = sigmoid(X @ w + b)
    return (X.T @ ((sigma - y) * sigma * (1 - sigma))) / n

How can I implement function with the same input for the piecewise formula from above?

Answer 1

I think the easiest way for this example is using condition multiplier as follows (not all elements in your function is defined in your question, so I am guessing them. However, you can fix it easily if it is not the case):

def loss_derivative(X, y, w):
    n = len(y)
    return X.T @ (((X @ w - y) / n) * ((X @ w) > 0))