gpCovGrads: Sparse objective function gradients wrt Covariance functions...
In alkalait/gptk: Gaussian Processes Tool-Kit

gives the gradients of the log likelihood with respect to the components of the sparse covariance (or the full covariance for the ftc case).

1	gpCovGrads(model, M)

`model`	the model for which the gradients are to be computed.
`M`	The training data for which the computation is to be made

`gK_uu`	the gradient of the likelihood with respect to the elements of K_uu (or in the case of the 'ftc' criterion the gradients with respect to the kernel).
`gK_uf`	the gradient of the likelihood with respect to the elements of K_uf.
`gLambda`	the gradient of the likelihood with respect to the diagonal term in the fitc approximation and the blocks of the pitc approximation.
`gBeta`	the gradient with respect to the beta term in the covariance structure.