BicopLike

class BicopLike(*args, **kwargs)

Contract for a bivariate (optionally conditional) pair copula.

A pair copula maps pseudo-observations u of shape (n, 2) (in the unit square, clamped to [1e-10, 1 - 1e-10]) to a density (pdf), a distribution (cdf), the two conditional distributions hfunc1(u) = P(U2 <= u2 | U1 = u1) / hfunc2(u) = P(U1 <= u1 | U2 = u2) and their inverses (hinv1 / hinv2, inverting in the second / first argument), plus a sampler (simulate). The optional x of shape (n, k) carries conditioning variables: a conditional copula reads them, an unconditional one ignores them.

The easy way to satisfy this contract is to subclass BicopBase, which supplies hinv1 / hinv2 (numerical inversion) and simulate on top of pdf / hfunc1 / hfunc2. pyvinecopulib.core.Bicop and pyvinecopulib.torch.TorchBicop are the reference implementations.

Methods

`cdf`(u[, x])	Pair-copula distribution `C(u)` at each observation.
`flip`()	Return the pair copula with its two arguments swapped.
`hfunc1`(u[, x])	First h-function `P(U2 <= u2 \| U1 = u1)`.
`hfunc2`(u[, x])	Second h-function `P(U1 <= u1 \| U2 = u2)`.
`hinv1`(u[, x])	Inverse of `hfunc1()` in its second argument.
`hinv2`(u[, x])	Inverse of `hfunc2()` in its first argument.
`pdf`(u[, x])	Pair-copula density `c(u)` at each observation.
`simulate`(n, *[, x, qrng, seeds])	Draw `n` samples from the pair copula.

See also

pyvinecopulib.core.BicopBase: Canonical partial implementation to subclass.
pyvinecopulib.core.Bicop: The reference pair copula.
VinecopLike: The vine-level evaluator contract.

Examples

A minimal independence pair on NumPy — implement only the three primitives and inherit hinv1 / hinv2 / simulate / loglik / plot / __repr__ from BicopBase:

import numpy as np
from pyvinecopulib.core import BicopBase

class Independence(BicopBase[np.ndarray]):
  def pdf(self, u, x=None):
    return np.ones(u.shape[0])

  def hfunc1(self, u, x=None):
    return u[:, 1]

  def hfunc2(self, u, x=None):
    return u[:, 0]

cop = Independence()
cop.hinv1(np.array([[0.3, 0.7]]))   # -> array([0.7]) (numerical inverse)

Methods

`__init__`
`cdf`	Pair-copula distribution `C(u)` at each observation.
`flip`	Return the pair copula with its two arguments swapped.
`hfunc1`	First h-function `P(U2 <= u2 \| U1 = u1)`.
`hfunc2`	Second h-function `P(U1 <= u1 \| U2 = u2)`.
`hinv1`	Inverse of `hfunc1()` in its second argument.
`hinv2`	Inverse of `hfunc2()` in its first argument.
`pdf`	Pair-copula density `c(u)` at each observation.
`simulate`	Draw `n` samples from the pair copula.