Verbleibende Zeit $1$:$04$:$31$

Assume you have a convolutional layer of $33$ filters with a total of $16$ filters and stride $2$ that is applied to inputs of dimensions $14,14,6$ and you want to initializethe weights of this conv$-$layer according to LeCun's Initialization scheme. Further assume that you want to use a normal distribution. What is the mean and variance of this normal distribution?

W$$hlen Sie eine Antwort:

a$.$ Mean: $1,$ Variance: $\frac{2}{3**3**6}$

b$.$ Mean: $0,$ Variance: $1$

c$.$ Mean: $0,$ Variance: $\frac{2}{3**3**16}$

d$.$ Mean: $1,$ Variance: $\frac{1}{3**3**6}$

e$.$ Mean: $0,$ Variance: $\frac{2}{3**3**6}$

f$.$ Mean: $0,$ Variance: $\frac{1}{3**3**6**16}$

g$.$ Mean: $0,$ Variance: $\frac{2}{3**3**6**16}$

h$.$ Mean: $0,$ Variance: $\frac{1}{3**3**6}$