TorchBicop.hinv1

TorchBicop.hinv1(u, x=None)

Inverts hfunc1 w.r.t. the second argument.

Given u = [u1, p], returns u2 such that H1(u1, u2) = p. With cache_integrals=True this is a single bilinear interpolation on the precomputed cache; otherwise each call inverts the piecewise-quadratic conditional cdf in closed form (exact inverse of the on-the-fly h-function).

Parameters:

uTensor, shape (n, 2), dtype float

Column 0 is u1; column 1 is the target probability p.

xTensor or None, optional

Conditioning variables, shape (n, k). Ignored by TorchBicop (an unconditional pair copula); accepted so the class satisfies the BicopLike contract.

Returns:

Tensor, shape (n,), dtype float

u2 values in [0, 1].