is it linear SARSA? #1
Description
Hello,
Thank you for sharing the code!
I have a question about what you describe MountainCar problem.
You said that 'Using SARSA with linear function approximation to solve the Mountain Car problem' and also 'The state samples were transformed into a higher dimensional space using an approximation of an RBF kernel allowing for a non-linear value function.'
I think it looks like 'non-linear' function approximation method more than 'linear' of that.
The reason why I think is that using non-linear function, in this case RBF kernel, to transform the state representation and than linear approximate the value function. Even taking linear approximation with state representation, you take non-linear function before linearization. Please let me know if I've missed out anything.
Thank you!