BatchNorm Forward

Problem

Normalize vector x to zero mean and unit variance, then scale and shift:

y[i] = gamma * (x[i] - mean) / sqrt(var + eps) + beta

where var is the biased (population) variance. If var + eps is 0, treat the normalized value as 0.

Hint

The epsilon inside the square root keeps the division safe when variance is small.

case 1

in: [[1,2,3],1,0,0]

expected: [-1.224744871391589,0,1.224744871391589]

case 2

in: [[1,1],1,0,0]

expected: [0,0]

case 3

in: [[1,2,3],2,1,0]

expected: [-1.4494897427831779,1,3.449489742783178]

case 4

in: [[0,2],1,1,1]

expected: [0.29289321881345254,1.7071067811865475]

Starter codePython

def batchnorm_forward(x, gamma, beta, eps):
    # Your code here
    pass

4 test cases · run your code against them in the editor