Visualizing hyperparameters with a robot arm

Using Gradient Descent to solve the Inverse Kinematics for a robot arm lets us visualize the impact that hyperparameters have on the optimization. Too high a learning rate and we start thrashing. Too low and we converge too fast. Too much momentum and we circle the target if we approach too quickly. Just the right amount of momentum more than doubles how fast we converge at the optimal solution.