Neural Tangent Kernel: The Linear Kernel Matrix
Beginner
The problem, from first principles
The neural tangent kernel (Jacot et al., 2018) says a very wide network trained by gradient descent behaves like a kernel method. The simplest kernel to start with is the linear one, a matrix of dot products between inputs.
From theory to code
Implement linear_kernel_matrix(X1, X2), returning the matrix of all pairwise dot products.
Constraints
- The output has shape
(n1, n2).
Hints
Hint 1
Multiply X1 by the transpose of X2.
Keywords
research-papersclassical-mlkernelsntk
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.