"Generalised diffusion map" provides extensions of the diffusion maps construction. For more details see Banisch, Trstanova, Bittracher, Klus and Koltai 2017.
TMDmap: target measure diffusion maps, generalizes diffusion maps (Coifman & Lafon)
to approximate the generator of a gradient flow Markov process
where the gradient terms are not slaved to the sampling density,
but are derived from a target probability measure that is chosen
by the user and known up to a normalization constant.
LKDmap: local kernel diffusion maps, allows to approximate the forward and backward
Fokker–Planck operators of a large class of Ito diffusions
FlowVisualisation contains an example of an application of LKDmap to flow analysis.