man advect/api/scipy/special

SPECIAL(3)Library CallsSPECIAL(3)

────────────────────────────────────────────────────────────────────────────────────────

Special Functions¶

Use these explicit Advect counterparts when a scipy.special function must remain differentiable or stageable. See the Scientific Python tutorial for a logsumexp gradient and the compatibility table for exact coverage.

special ¶

Traceable high-value counterparts to scipy.special.

gammaln(
    x: object, /, out: object = None, **kwargs: object
) -> object

Compute the logarithm of the absolute gamma function.

digamma(
    x: object, /, out: object = None, **kwargs: object
) -> object

Compute the logarithmic derivative of the gamma function.

polygamma(n: object, x: object) -> object

Compute the n-th derivative of digamma with SciPy broadcasting.

erf(
    x: object, /, out: object = None, **kwargs: object
) -> object

Compute the error function.

erfc(
    x: object, /, out: object = None, **kwargs: object
) -> object

Compute the complementary error function.

erfcx(
    x: object, /, out: object = None, **kwargs: object
) -> object

Compute the scaled complementary error function.

erfinv(
    y: object, /, out: object = None, **kwargs: object
) -> object

Compute the inverse error function.

expit(
    x: object, /, out: object = None, **kwargs: object
) -> object

Compute the logistic sigmoid.

log_expit(
    x: object, /, out: object = None, **kwargs: object
) -> object

Compute the logarithm of the logistic sigmoid.

ndtr(
    x: object, /, out: object = None, **kwargs: object
) -> object

Compute the standard normal cumulative distribution function.

log_ndtr(
    x: object, /, out: object = None, **kwargs: object
) -> object

Compute the logarithm of the standard normal cumulative distribution.

ndtri(
    p: object, /, out: object = None, **kwargs: object
) -> object

Compute the inverse standard normal cumulative distribution.

logsumexp(
    a: object,
    axis: object = None,
    b: object = None,
    keepdims: bool = False,
    return_sign: bool = False,
) -> object

Compute SciPy-compatible weighted, optionally signed log-sum-exp.

softmax(x: object, axis: object = None) -> object

Compute the softmax function along axis.

log_softmax(x: object, axis: object = None) -> object

Compute the logarithm of the softmax function along axis.

$ █