torchelie.distributions¶
-
class
torchelie.distributions.GaussianMixture(weights: torch.Tensor, locs: torch.Tensor, scales: torch.Tensor)¶ Mixture of gaussian distributions. Each tensor contains an additional dimension with number of distributions elements.
- Parameters
weights (tensor) – un-normalized weights of distributions
loc (tensor) – mean of the distributions
scale (tensor) – scale of the distributions
dim (int) – dimension reprenseting the various distributions, that will weighted and averaged on.
-
log_prob(x: torch.Tensor) → torch.Tensor¶
-
property
mean¶
-
class
torchelie.distributions.Logistic(loc: torch.Tensor, scale: torch.Tensor)¶ Logistic distribution
- Parameters
loc (tensor) – mean of the distribution
scale (tensor) – scale of the distribution
-
class
torchelie.distributions.LogisticMixture(weights, locs, scales, dim)¶ Mixture of Logistic distributions. Each tensor contains an additional dimension with number of distributions elements.
- Parameters
weights (tensor) – un-normalized weights of distributions
loc (tensor) – mean of the distributions
scale (tensor) – scale of the distributions
dim (int) – dimension reprenseting the various distributions, that will weighted and averaged on.
-
log_prob(x: torch.Tensor) → torch.Tensor¶
-
property
mean¶
-
torchelie.distributions.parameterized_truncated_normal(uniform: torch.Tensor, mu: float, sigma: float, a: float, b: float) → torch.Tensor¶ Experimental
Warning
parameterized_truncated_normal() is experimental, and may change or be deleted soon if not already broken
.
-
torchelie.distributions.sample_truncated_normal(*shape, cutoff: float = 2)¶ Experimental
Warning
sample_truncated_normal() is experimental, and may change or be deleted soon if not already broken
.
-
torchelie.distributions.truncated_normal(uniform: torch.Tensor, a: float, b: float) → torch.Tensor¶ Experimental
Warning
truncated_normal() is experimental, and may change or be deleted soon if not already broken
.