深入浅出Pytorch函数——torch.nn.init.xavier_normal_

torch.nn.init模块中的所有函数都用于初始化神经网络参数，因此它们都在torc.no_grad()模式下运行，autograd不会将其考虑在内。

根据Glorot, X.和Bengio, Y.在《Understanding the difficulty of training deep feedforward neural networks》中描述的方法，用一个正态分布生成值，填充输入的张量或变量。结果张量中的值采样自 N ( 0 , std 2 ) N(0, \text{std}^2) N(0,std2)的正态分布，其中标准差：
std = gain × 2 fan_in + fan_put \text{std}=\text{gain}\times\sqrt{\frac{2}{\text{fan\_in}+\text{fan\_put}}} std=gain×fan_in+fan_put2

这种方法也被称为Glorot initialisation。

语法

复制代码

torch.nn.init.xavier_normal_(tensor, gain=1.0)

参数

tensor：[Tensor] 一个 N N N维张量torch.Tensor
gain ：[float] 可选的缩放因子

返回值

一个torch.Tensor且参数tensor也会更新

实例

复制代码

w = torch.empty(3, 5)
nn.init.xavier_normal_(w)

函数实现

复制代码

def xavier_normal_(tensor: Tensor, gain: float = 1.) -> Tensor:
    r"""Fills the input `Tensor` with values according to the method
    described in `Understanding the difficulty of training deep feedforward
    neural networks` - Glorot, X. & Bengio, Y. (2010), using a normal
    distribution. The resulting tensor will have values sampled from
    :math:`\mathcal{N}(0, \text{std}^2)` where

    .. math::
        \text{std} = \text{gain} \times \sqrt{\frac{2}{\text{fan\_in} + \text{fan\_out}}}

    Also known as Glorot initialization.

    Args:
        tensor: an n-dimensional `torch.Tensor`
        gain: an optional scaling factor

    Examples:
        >>> w = torch.empty(3, 5)
        >>> nn.init.xavier_normal_(w)
    """
    fan_in, fan_out = _calculate_fan_in_and_fan_out(tensor)
    std = gain * math.sqrt(2.0 / float(fan_in + fan_out))

    return _no_grad_normal_(tensor, 0., std)