Third step: adding noise to the data

The last step needs only three parameters related to the addition of noise to expression data. In fact, the procedure consider two types of noise: the first one affects the classification of the tissues and is characterized by the parameter $p$, which represents the probability that a wrong label is assigned. The second one is noise added to gene expression data obtained from the second step. In this case, with a fixed probability $e$, a real value, chosen according to a normal distribution with mean 0 and standard deviation $(x_{\max}-x_{\min})/\nu$, is added to each element of the matrix. It follows that values for the parameters $p$, $e$ and $\nu$ must be selected to perform the final step of the procedure.