agedi.utils.truncated_normal¶

BSD 3-Clause License

Copyright (c) 2020, Anton Obukhov All rights reserved.

https://github.com/toshas/torch_truncnorm

Attributes¶

`CONST_SQRT_2`
`CONST_INV_SQRT_2PI`
`CONST_INV_SQRT_2`
`CONST_LOG_INV_SQRT_2PI`
`CONST_LOG_SQRT_2PI_E`

Classes¶

`TruncatedStandardNormal`	Truncated Standard Normal distribution
`TruncatedNormal`	Truncated Normal distribution

Module Contents¶

agedi.utils.truncated_normal.CONST_SQRT_2¶

agedi.utils.truncated_normal.CONST_INV_SQRT_2PI¶

agedi.utils.truncated_normal.CONST_INV_SQRT_2¶

agedi.utils.truncated_normal.CONST_LOG_INV_SQRT_2PI¶

agedi.utils.truncated_normal.CONST_LOG_SQRT_2PI_E¶

class agedi.utils.truncated_normal.TruncatedStandardNormal(a: numbers.Number | torch.Tensor, b: numbers.Number | torch.Tensor, validate_args: bool | None = None)¶

Bases: torch.distributions.Distribution

Truncated Standard Normal distribution https://people.sc.fsu.edu/~jburkardt/presentations/truncated_normal.pdf

arg_constraints¶

has_rsample = True¶

_dtype_min_gt_0¶

_dtype_max_lt_1¶

_little_phi_a¶

_little_phi_b¶

_big_phi_a¶

_big_phi_b¶

_Z¶

_log_Z¶

_lpbb_m_lpaa_d_Z¶

_mean¶

_variance¶

_entropy¶

support()¶: Return the support interval [a, b].

property mean: torch.Tensor¶: Mean of the truncated standard normal distribution.

property variance: torch.Tensor¶: Variance of the truncated standard normal distribution.

property entropy: torch.Tensor¶: Differential entropy of the truncated standard normal distribution.

property auc: torch.Tensor¶: Normalisation constant Z = Φ(b) − Φ(a).

static _little_phi(x: torch.Tensor) → torch.Tensor¶: Standard normal probability density function φ(x).

static _big_phi(x: torch.Tensor) → torch.Tensor¶: Standard normal cumulative distribution function Φ(x).

static _inv_big_phi(x: torch.Tensor) → torch.Tensor¶: Inverse of the standard normal CDF Φ⁻¹(x).

cdf(value: torch.Tensor) → torch.Tensor¶: Cumulative distribution function evaluated at value.

icdf(value: torch.Tensor) → torch.Tensor¶: Inverse CDF (quantile function) evaluated at value.

log_prob(value: torch.Tensor) → torch.Tensor¶: Log probability density evaluated at value.

rsample(sample_shape: torch.Size = torch.Size()) → torch.Tensor¶: Draw a re-parameterised sample of the given shape.

class agedi.utils.truncated_normal.TruncatedNormal(loc: numbers.Number | torch.Tensor, scale: numbers.Number | torch.Tensor, a: numbers.Number | torch.Tensor, b: numbers.Number | torch.Tensor, validate_args: bool | None = None)¶

Bases: TruncatedStandardNormal

Truncated Normal distribution https://people.sc.fsu.edu/~jburkardt/presentations/truncated_normal.pdf

has_rsample = True¶

_log_scale¶

_mean¶

_variance¶

_to_std_rv(value: torch.Tensor) → torch.Tensor¶: Standardise value to the standard (zero-mean, unit-variance) domain.

_from_std_rv(value: torch.Tensor) → torch.Tensor¶: Map value from the standard domain back to the original (loc/scale) domain.

cdf(value: torch.Tensor) → torch.Tensor¶: Cumulative distribution function evaluated at value.

icdf(value: torch.Tensor) → torch.Tensor¶: Inverse CDF (quantile function) evaluated at value.

log_prob(value: torch.Tensor) → torch.Tensor¶: Log probability density evaluated at value.