Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It only takes a minute to sign up.
Sign up to join this community
Suppose we have feature transformation $\Phi(x) = [1, x_1, x_2, x_1x_2]$. Now we want to find the kernel corresponding to $\Phi$.
What I have done is using kernel decomposition, we have: $$ K(x, y) = \Phi(x) .\Phi(y)\\ K(x, y) = 1.1 + x_1y_1 + x_2y_2 + x_1x_2y_1y_2 \\ K(x, y) = 1 + \sum_{i=1}^{N}x_iy_i + \prod_{i=1}^{N}x_iy_i $$
$x, y$ are in $\Bbb{R}^2$
Is that kernel valid?
Yes, it is valid.
Kernel is just an inner product in the feature space. In fact, here you have explicitly started from the feature transformation and we understand the feature transformation.
