Distributions

[krzjoa]

A checklist of distributions we'd like to port from PyTorch:

• [x] Bernoulli
• [ ] Beta
• [ ] Binomial
• [x] Categorical
• [ ] Cauchy
• [x] Chi2
• [ ] ContinuousBernoulli
• [ ] Dirichlet
• [ ] Exponential
• [ ] FisherSnedecor
• [x] Gamma
• [ ] Geometric
• [ ] Gumbel
• [ ] HalfCauchy
• [ ] HalfNormal
• [ ] Independent
• [ ] Laplace
• [ ] LogNormal
• [ ] LowRankMultivariateNormal
• [x] MixtureSameFamily
• [ ] Multinomial
• [x] MultivariateNormal
• [ ] NegativeBinomial
• [x] Normal
• [ ] OneHotCategorical
• [ ] Pareto
• [x] Poisson
• [ ] RelaxedBernoulli
• [ ] LogitRelaxedBernoulli
• [ ] RelaxedOneHotCategorical
• [ ] StudentT
• [ ] TransformedDistribution
• [ ] Uniform
• [ ] VonMises
• [ ] Weibull

[skiptoniam]

@dfalbel I was keen to use a negative binomial distribution. Any pointers on how to set one up in torch?