BatchNorm Fold Bias

Problem

Fold batch normalization into the convolution bias.

For each output channel o return round((bias[o] - mean[o]) * gamma[o] / sqrt(var[o] + eps) + beta[o], 4).

case 1

in: [[1,2],[0,1],[1,2],[1,4],[0,1],0]

expected: [1,2]

case 2

in: [[0],[1],[1],[1],[0],0.000001]

expected: [-1]

case 3

in: [[2,4],[0,2],[1,1],[4,4],[0,0],0]

expected: [1,1]

Starter codePython

def fold_conv_bias(bias, mean, gamma, var, beta, eps):
    # Your code here
    pass

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