Options

# Q: How to cascade Optimize Parameters (Grid)?

I use Grid Parameter Optimization for my SVR processes (C, gamma). How can I do a cascaded loop?

e.g. I optimize for C = 50 and gamma loops between 2^n where n = [-8, 8]

Suppose optimal solution is for n = -5

Now, I want to make a process that will:

1. take from a log values corresponding to optimum of 5, such as:

2^n, where n = -5,5, -5,4, -5,3, -5,2, -5,1, -5,0, -4,9, -... -4,5

and optimize gamma for these to get optimum solution in the second step.

Afterwards I want to take the best gamma and optimize C on it.

Final thing is that all this is done in a loop. My question is how to:

-update grid parameter optimization parameters based on the optimum value of the last run, and doing a new run?

e.g. I optimize for C = 50 and gamma loops between 2^n where n = [-8, 8]

Suppose optimal solution is for n = -5

Now, I want to make a process that will:

1. take from a log values corresponding to optimum of 5, such as:

2^n, where n = -5,5, -5,4, -5,3, -5,2, -5,1, -5,0, -4,9, -... -4,5

and optimize gamma for these to get optimum solution in the second step.

Afterwards I want to take the best gamma and optimize C on it.

Final thing is that all this is done in a loop. My question is how to:

-update grid parameter optimization parameters based on the optimum value of the last run, and doing a new run?

0