Click4Ai

97.

Easy

Implement the **Swish** activation function and its derivative.

Swish:

swish(x) = x * sigmoid(beta * x)

Derivative:

swish'(x) = sigmoid(beta*x) + beta * x * sigmoid(beta*x) * (1 - sigmoid(beta*x))

= swish(x) + sigmoid(beta*x) * (beta - beta * swish(x))

Example:

swish(0) → 0.0

swish(1) → 0.7311

swish(-1) → -0.2689

**Explanation:** Swish was discovered by Google Brain using automated search. Unlike ReLU, it's smooth and non-monotonic (slightly negative for small negative inputs). It often outperforms ReLU in deep networks. With beta=1, it's equivalent to SiLU.

Constraints:

  • Default beta = 1.0
  • Input can be a scalar or numpy array

    • Test Cases

    Test Case 1
    Input: x=0, beta=1
    Expected: swish=0.0
    Test Case 2
    Input: x=[-2,-1,0,1,2], beta=1
    Expected: swish=[-0.2384,-0.2689,0,0.7311,1.7616]
    Test Case 3
    Input: x=1, beta=1
    Expected: swish=0.7311
    + 2 hidden test cases