In SVM, there is a bias term. But looks to me there are very few discussions on the physical meanings of this term. Why should we have that? How does this term affect the model?
This is a paper you probably should read:
Poggio, T., S. Mukherjee, R. Rifkin, A. Rakhlin and A. Verri. b, CBCL Paper #198/AI Memo #2001-011, Massachusetts Institute of Technology, Cambridge, MA, July 2001. (PostScript)
I rather doubt there is a paper with a shorter title!