Skip to content

penalized_zero_inflated_truncated_gaussian

Regularized zero-inflated truncated Gaussian marginal model.

PenalizedZeroInflatedTruncatedGaussian(formula, *, prior_alpha=1.1, penalties=(), device=None)

A regularized zero-inflated truncated Gaussian marginal

The penalty P is added to the training loss, together with a Beta(prior_alpha, prior_alpha) prior term already present in :class:ZeroInflatedTruncatedGaussian.

Parameters:

Name Type Description Default
formula dict of str to str, or str
required
prior_alpha float

See :class:ZeroInflatedTruncatedGaussian.

1.1
penalties sequence of CoefficientPenalty

See :mod:scdesigner.distributions.penalties.

()
device device or str
None

Examples:

>>> from scdesigner.distributions.penalties import CoefficientPenalty, QuadraticPenalty
>>> penalty = CoefficientPenalty("mean", ["x"], [1.0], QuadraticPenalty(lam=0.1))
>>> model = PenalizedZeroInflatedTruncatedGaussian({"mean": "~ x"}, penalties=[penalty])
>>> type(model).__name__
'ZeroInflatedTruncatedGaussian'
Source code in scdesigner/src/scdesigner/distributions/penalized_zero_inflated_truncated_gaussian.py
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
def PenalizedZeroInflatedTruncatedGaussian(
    formula: Union[Dict[str, str], str],
    *,
    prior_alpha: float = 1.1,
    penalties: Sequence[CoefficientPenalty] = (),
    device=None,
) -> ZeroInflatedTruncatedGaussian:
    """A regularized zero-inflated truncated Gaussian marginal

    The penalty ``P`` is added to the training loss, together with a
    Beta(``prior_alpha``, ``prior_alpha``) prior term already present in
    :class:`ZeroInflatedTruncatedGaussian`.

    Parameters
    ----------
    formula : dict of str to str, or str
    prior_alpha : float
        See :class:`ZeroInflatedTruncatedGaussian`.
    penalties : sequence of CoefficientPenalty
        See :mod:`scdesigner.distributions.penalties`.
    device : torch.device or str, optional

    Examples
    --------
    >>> from scdesigner.distributions.penalties import CoefficientPenalty, QuadraticPenalty
    >>> penalty = CoefficientPenalty("mean", ["x"], [1.0], QuadraticPenalty(lam=0.1))
    >>> model = PenalizedZeroInflatedTruncatedGaussian({"mean": "~ x"}, penalties=[penalty])
    >>> type(model).__name__
    'ZeroInflatedTruncatedGaussian'
    """
    return ZeroInflatedTruncatedGaussian(
        formula, prior_alpha=prior_alpha, penalties=penalties, device=device
    )