I am trying to implement the SDCA algorithm and the pseudocode is shown in the image below. My question is how to solve the 1D minimization problem. I am using a logistic loss function with the usual regularization term.
I know that the gradient is not available in the closed form for that problem and hence I need to use methods like golden section or bisection methods but both of them require an interval and I am not sure how to supply the interval. Any ideas how to solve this problem?