Consider the regularized SVM approach: \begin{equation} g (\alpha) = \sum_{i=1}^{n} \alpha_i - \frac{1}{2}\sum_{i=1}^{n} \sum_{j=1}^{n} \alpha_i \alpha_j y_i y_j \langle \phi(x_i), \phi(x_j)\rangle_H \end{equation} under the constraints $0\leq \alpha_i \leq C$ for $i=1, \dots, n$ and $\sum_{i=1}^{n} \alpha_i y_i = 0$, where $\phi$ is some feature-map into a Hilbert space $H$.

What is the easiest method to solve this? Is something like the SOM-Algorithm necessarily needed?

If one tries to solve this analytically (setting the gradient to zero and solving the system of n linear equations), it doesn't work because the parameter $C$ does not appear...


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.