DeepARLikelihood

Enums · AiDotNet.TimeSeries

The likelihood of a DeepAR network’s observations (Salinas et al. 2020, §3.1).

Fields

Field Summary
DeepARLikelihood Gaussian Gaussian for real-valued data: μ = wᵀh + b, σ = softplus(wᵀh + b), both multiplied by ν.
DeepARLikelihood NegativeBinomial Negative binomial for count data: μ = ν·softplus(·), α = softplus(·)/√ν, Var = μ + μ²α.
DeepARLikelihood StudentT Student-t, the GluonTS reference’s default: df = 2 + softplus(·), loc and softplus scale times ν.
int value__