Gaussian Process package based on data augmentation, sparsity and natural gradients
For breaking points see #106.
Mostly SVGP
and OnlineSVGP
are affected:
SVGP(X, y, kernel, likelihood, inference, n_ind_points)
becomes SVGP(kernel, likelihood, inference, Z)
and train!(model)
becomes train!(model, X, y)
m = train!(m::OnlineSVGP, x, y)
becomes m, new_state = train!(m, x, y, state)
Merged pull requests:
Merged pull requests: