Distributions
- class multiviewae.base.distributions.Default(**kwargs)[source]
Artificial distribution designed for data with unspecified distribution. Used so that log_likelihood and _sample methods can be called by model class. :param x: List of input data. :type x: list
- class multiviewae.base.distributions.Categorical(**kwargs)[source]
Artificial distribution designed for categorical data. Used so that log_likelihood and _sample methods can be called by model class. :param x: List of input data. :type x: list
- class multiviewae.base.distributions.Normal(**kwargs)[source]
Univariate normal distribution. Inherits from torch.distributions.Normal.
- Parameters
loc (int, torch.Tensor) – Mean of distribution.
scale (int, torch.Tensor) – Standard deviation of distribution.
- property variance
Returns the variance of the distribution.
- sparse_kl_divergence()[source]
Implementation from: https://github.com/senya-ashukha/variational-dropout-sparsifies-dnn/blob/master/KL%20approximation.ipynb
- class multiviewae.base.distributions.MultivariateNormal(**kwargs)[source]
Multivariate normal distribution with diagonal covariance matrix. Inherits from torch.distributions.multivariate_normal.MultivariateNormal.
- Parameters
loc (list, torch.Tensor) – Mean of distribution.
scale (int, torch.Tensor) – Standard deviation of distribution.
- property variance
Returns the variance of the distribution.
- class multiviewae.base.distributions.Bernoulli(**kwargs)[source]
Bernoulli distribution. Inherits from torch.distributions.Bernoulli. :param x: List of input data. :type x: list
- class multiviewae.base.distributions.ApproxBernoulli(**kwargs)[source]
Artificial distribution designed for (approximately) Bernoulli distributed data. The data isn’t restricted to bernoulli distribution, this class is designed as a wrapper for the log_likelihood() method which is required for the multiview models.
- Parameters
x (list) – List of input data.