Differentiating Gaussian Processes

Only the k(x * , X) term depends on the test point x * , therefore to calculate the slope of the posterior mean we just need to differentiate the kernel. For the squared exponential covariance function the derivative of the kernel between x * and a training point x i is,