m.weight.data.normal_(0, math.sqrt(2. / n)) what does it mean?

[zhang405744522]

[JiaRenChang]

@zhang405744522 It is weight initialization. Please refer to Kaiming He 2015 paper

[zhang405744522]

thanks for your response so fast.
I also got the papaer <Understanding the difficulty of training deep feedforward neural networks>

[zhang405744522]

Understanding the difficulty of training deep feedforward neural networks

[niraj2477]

Can anyone please tell me , how can I initialize it in keras

[zgh1349786296]

    for m in self.modules():
        if isinstance(m, nn.Conv2d):
            n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
            m.weight.data.normal_(0, math.sqrt(2. / n))
        elif isinstance(m, nn.BatchNorm2d):
            m.weight.data.fill_(1)
            m.bias.data.zero_()