Hi guys, I just finished watching this amazing lecture on Support Vector Machines by Prof. Patrick Winston. I was amazed by the applications of multiple kernel functions as he has mentioned in his lecture.
I also checked out the scikit-learn documentation of SVM and they have mentioned something about custom kernel functions. I am putting the link up here.
Check out the section 1.4.6.
I wonder if you have implemented some custom kernel function in any of your projects and got better results than using already defined kernels.
Also, is there any theory about which kernel should be used when?