He and Glorot Initialization Scales
~10 mincode completion
Implement init_scales(n_in, n_out, sigma_w, depth) returning the array [he_std, glorot_std, gain, gain ** depth].
Where this shows up
~10 min
••••••••••••••••
8 employers weight this skill
4 frontier labs, 2 big tech firms, 1 autonomy company, 1 enterprise vendor. Top match scores 93.
Python
import numpy as np
def init_scales(n_in, n_out, sigma_w, depth):
"""
Initialization scales and the signal gain they imply.
Args:
n_in: inputs to each unit
n_out: units in the layer
sigma_w: standard deviation the weights are actually drawn with
depth: number of identical ReLU layers stacked
Returns:
array [he_std, glorot_std, gain, gain ** depth]
"""
# YOUR CODE HERE
pass