A kernel that describes how very wide neural networks change during training.
NTK is like a toy train track for a huge neural net. The engine moves, but the rails decide its path.
It helps researchers study ultra-wide networks. You meet it in theory papers on deep learning and Kernel Methods.
Kernel Method
NTK connects ultra-wide neural network training to kernel regression.
Neural-network
It describes how a neural network trains near its starting point.
Overparameterization
The NTK approximation works more easily when the network is very wide.
Gradient Descent
It describes how gradient descent changes the model output.