In section 8.3 on support vector regression, you have equation 8.19 for the dual optimisation problem for epsilon-insensitive support vector regression. You say that it is similar to the derivation for support vector classification in the appendix.
I went through the derivation, using the appendix as a guide, but I'm unsure of how to get into the form in 8.19. I've found the form that I derive given in various sources online, but can't find any sources that have the same form as you give.
Could you tell me the step needed to get into the form given in 8.19. Also what is the advantage of it being in this form? The equation I derive can immediately be put into numerical quadratic optimisation solvers. I don't see how 8.19 can be. Is it still quadratic?
In section 8.3 on support vector regression, you have equation 8.19 for the dual optimisation problem for epsilon-insensitive support vector regression. You say that it is similar to the derivation for support vector classification in the appendix.
I went through the derivation, using the appendix as a guide, but I'm unsure of how to get into the form in 8.19. I've found the form that I derive given in various sources online, but can't find any sources that have the same form as you give.
Could you tell me the step needed to get into the form given in 8.19. Also what is the advantage of it being in this form? The equation I derive can immediately be put into numerical quadratic optimisation solvers. I don't see how 8.19 can be. Is it still quadratic?
I've attached my notes on the derivation: SVR.pdf
Thank you,
Owen