More intense Gaussian Noise

I want to apply more intense noise to the input data. Is it possible?
The current Noise in the model

class GaussianNoise(nn.Module):
def init(self, sigma):
super(GaussianNoise, self).init()
self.sigma = sigma

def forward(self, input):
    if self.training:
        noise = Variable(input.data.new(input.size()).normal_(std=self.sigma))
        return input + noise
    else:
        return input

Hi KanZa!

Can’t you just use a larger value of sigma when you instantiate your
GaussianNoise object?

As an aside, this code looks rather out of date. torch.autograd.Variable
has been deprecated for some time now.

Best.

K. Frank

Thank you for your suggestion.
Can you please elaborate on your suggestion, I tried by taking larger Sigma values.
Any other current Gaussian Noise method you would like to suggest?

Hi KanZa!

I assume that you are saying that using larger values for sigma did not
successfully “apply more intense noise.”

Could you post a short, self-contained, runnable script that shows what
you tried, together with its output, and explain what you were looking for
that you didn’t get?

Best.

K. Frank