Kernel Logistic Regression

Let’s take the logistic regression algorithm with gradient descent. The kernel form for the update rule is

\[ a(t + 1) = a(t) + \alpha (y-s(Ka)) \]

The model is then,

\[ \hat y (z) = s\left(\sum_{i=1}^n a_j k(X_j, z)\right) \]