VHRanger on Every Model Learned by Gradient Descent Is Approximately a Kernel Machine · Hackerly

Comment on Every Model Learned by Gradient Descent Is Approximately a Kernel Machine parent

Also, cubic splines don't extrapolate like the other mentioned models