Distributions

class torchelie.distributions.GaussianMixture(weights, locs, scales)

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)
mean
class torchelie.distributions.Logistic(loc, scale)

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)
mean