Suppose I have some data vector, $\mathbf{d}$ with length $N$. I want to add "Gaussian noise" to the data. My understanding is that there are two ways to do this.

1) Relative noise percentage

$$\mathbf{d_n} = \mathbf{d}+\epsilon\mathbf{dr}$$

where $\mathbf{d_n}$ is the noisy data, $\epsilon$ is the error fraction and $\mathbf{r}$ is a vector of numbers drawn from the Gaussian distribution.

2) "White" noise

$$\mathbf{d_n} = \mathbf{d}+\xi\mathbf{r}$$

where $\xi$ is the "noise floor".

In the first scenario, the added noise is dependent on the data itself (i.e. a large data value will likely have a larger noise term than a smaller data value). In the second scenario, the noise is independent of the data.

Which is "better"?

Why choose one method over the other?

Any information is appreciated.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.