• Typically, these two parameters are tuned together. Since making the learning rates smaller, will require more trees to maintain model complexity.