[qPM9l3XJrF]

I think you'd be better off using a Gaussian process than reinforcement learning