According to this link, nu specifies the nu parameter for the one-class SVM model. The nu parameter is both a lower bound for the number of samples that are support vectors and an upper bound for the number of samples that are on the wrong side of the hyperplane. The default is 0.1. The nu parameter must be in the range [0,1].
I do not understand it very well, graphically as it would be?
Of theory I find the rank to which nu belongs, but what is the goal of nu for oneClass SVM, is nu equal to C?
Best Answer
The goal of $\nu$ is similar to that of $C$. The parameter $\nu$ is there to finetune the trade-off between overfitting and generalization. The problem with $C$ is that it is positive and unbounded which makes it difficult to choose optimally during cross-validation. Thus, the range $[0,1]$ makes the regularization more interpretable in terms of $\nu$ but empirically, people have found it more difficult to optimize in terms of $\nu$.
The relationship between $\nu$ and $C$ is given by the formula $\nu= \frac{A+B}{C}$ where $A$ and $B$ are some constants that are not easy to calculate.
The explanation/intuition about the interpretation of $\nu$ given in the link you've mentioned is pretty accurate. $\nu$ is upper bounded by the fraction of outliers and lower bounded by the fraction of support vectors. Just consider that for default value $0.1$, atmost $10\%$ of the training samples are allowed to be wrongly classified or can be considered as outliers by the decision boundary. And atleast $10\%$ of your training samples will act as support vectors (points on the decision boundary).