Xavier Initialization, Kaming Initialization 权重初始化

Xavier Initialization (\(\text{Understanding the difficulty of training deep feedforward neural networks}\))

\(\text{Goal: The goal of Xavier Initialization is to initialize the weights such that the variance of the activations are the same across every layer. This constant variance helps prevent the gradient from exploding or vanishing.}\)

\(\text{Suppose we have an input: } X \text{ with }n\text{ components and a linear neuron with random weights }W\text{, we can write: }\)

\[\begin{align} Y = W_1X_1+W_2X_2+...+W_nX_n \end{align} \]

\(\text{Consider one term: }W_iX_i,\text{ the variance is }\)

\[\begin{align} Var(W_iX_i) = \mathbb{E}(X_i)^2Var(W_i)+\mathbb{E}(W_i)^2Var(X_i)+Var(X_i)Var(W_i) \end{align} \]

\(\text{If we assume that the variables are zero-mean, which can be simplified as:}\)

\[\begin{align} Var(W_iX_i) = Var(X_i)Var(W_i) \end{align} \]

\(\text{Furthermore, if we assume that }X_i,W_i\text{ are i.i.d, then we can get: }\)

\[\begin{align} Var(Y) &= Var(\sum_iW_iX_i)\\ &=n \cdot Var(W_i)Var(X_i) \end{align} \]

\(\large \text{From now we can observe, the output's variance }Var(Y)\text{ is also related to the input's variance, but scaled by }n\cdot Var(W_i). \text{ Therefore, if we want to control our output's variance (e.g. as the same as the input's variance), we need such conditions:}\)

\[Var(W_i) = \frac{1}{n} = \frac{1}{fan\_in} \]

\(\text{You can see this is actually the Forward process, if we consider the BackPropagation, we might need:}\)

\[Var(W_i) =\frac{1}{fan\_out} \]

\(\text{However, in real NN architectures, it's not common to have same number of inputs and outputs neurons. To compromise this, we take the average of two:}\)

\[\begin{align} Var(W_i) = \frac{2}{fan\_in+fan\_out} \end{align} \]

\(\text{In summary, the assumptions we need to derive our results:}\)
\(\text{I. }W,X\text{ are zero-mean}\)
\(\text{II. }W,X\text{ are I.I.D}\)
\(\text{III. Biases are initialized as zeros}\)
\(\text{IV. We use the }\tanh()\) \(\text{ activation function, which is approximately as linear in small inputs: }Var(a^{[l]})\approx Var(z^{[l]})\)

Kaiming Initialization

网上很多博客解释的并不好，大多只是介绍结果。这里参考论文来推导一下：
\(\text{From Xavier's results, we know that: }Var[y_l] = n_lVar[w_lx_l]. \text{ Then if }w_l \text{ has zero-mean, we can further obtain: }\)

\[\begin{align} Var[y_l] &= n_lVar[w_l](\mathbb{E}(x_l)^2+Var[x_l])\\ &=n_lVar[w_l]\mathbb{E}(x_l^2) \end{align} \]

对于\(ReLU\)函数，\(x_l = \max\{0,y_{l-1}\}\)，因此\(\mathbb{E}(x_l^2)\neq Var(x_l)\).
假设\(w_{l-1}\)关于\(0\)对称分布，\(b_{l-1}=0\). 现考虑\(\mathbb{E}(x_l^2)=\mathbb{E}(\max(0,y_{l-1})^2)\). 因此 \(y_{l-1}\)也关于\(0\)对称分布。注意到：

\[\begin{align} \mathbb{P}(y_{l-1}>0)&=\mathbb{P}(w_{l-1}x_{l-1}>0)\\ &=\mathbb{P}[(w_{l-1}>0 \& x_{l-1}>0) \|(w_{l-1}<0 \& x_{l-1}<0)]\\ &=\mathbb{P}(x_{l-1}>0)\mathbb{P}(w_{l-1}>0)+\mathbb{P}(x_{l-1}<0)\mathbb{P}(w_{l-1}<0)\\ &=\frac{1}{2}\mathbb{P}(x_{l-1}>0)+\frac{1}{2}\mathbb{P}(x_{l-1}<0)\\ &=1/2 \end{align} \]

进一步：

\[\begin{align} \mathbb{E}[x_l^2] &=\mathbb{E}[\max(0,y_{l-1})^2]\\ &=\frac{1}{2}\mathbb{E}[y_{l-1}^2]\\ &=\frac{1}{2}Var[y_{l-1}] \end{align} \]

因此：

\[\begin{align} Var[y_l] &= \frac{1}{2}n_lVar[w_l]Var[y_{l-1}] \end{align} \]

\(\text{Consider all }L\text{ layers: }\)

\[\begin{align} Var[y_l] = Var[y_1](\prod_{l=2}^L\frac{1}{2}n_lVar[w_l]) \end{align} \]

A proper initialization method should avoid reducing or magnifying the magnitudes of input signals exponentially.

从这可以很明显地发现，一个充分条件就是：

\[\begin{align} \frac{1}{2}n_lVar[w_l]=1 \end{align} \]

This leads to a zero-mean Gaussian distribution whose standard deviation (std) is \(\sqrt{2/n_l}\). This is our way of initialization. We also initialize \(b = 0\).

\(\text{The difference between this Kaiming initialization and the Xavier one is again the 1/2 that comes from the ReLU activation function.}\)