The GradientJumps currently implement a coordinate transformation, through the functions: forward, backward, func_grad_white, draw_momenta, loghamiltonian, and posmom_inprod.
In principle, this is unnecessary, if only the functions draw_momenta, loghamiltonian, and posmom_inprod are modified properly. It should be a simple matter of using the right version of cov_cf/cov_cfi and transposes, but somehow I did not get it to work.
The GradientJumps work fine like this, it is just that it looks clumsy.
The GradientJumps currently implement a coordinate transformation, through the functions: forward, backward, func_grad_white, draw_momenta, loghamiltonian, and posmom_inprod.
In principle, this is unnecessary, if only the functions draw_momenta, loghamiltonian, and posmom_inprod are modified properly. It should be a simple matter of using the right version of cov_cf/cov_cfi and transposes, but somehow I did not get it to work.
The GradientJumps work fine like this, it is just that it looks clumsy.