neural_operators.prior

class neural_operators.prior.priorSampler.PriorSampler(V, a, c, seed=0)

Bases: object

Gaussian random field prior via the elliptic SPDE precision operator L = -a div(b grad(.)) + c, covariance C = L^-2. a and c set correlation length and marginal variance; b (default 1, see set_diffusivity) lets correlation length vary in space.

Call the instance to draw a sample; logPrior(m) evaluates the (unnormalized) log-density at a vertex-ordered field.

assemble()

compute_mean(m)

empty_sample()

function_to_vector(u_fn, u_vec=None)

function_to_vertex(u_fn, u_vv=None)

logPrior(m)

set_diffusivity(diffusion)

vector_to_function(u_vec, u_fn=None)

vertex_to_function(u_vv, u_fn=None)