Expand description
Maximum-likelihood estimation: the generic MLE framework plus the concrete per-distribution likelihood models built on top of it.
A parametric model implements LogLikelihood and defers optimization to
fit_mle, which maximizes ℓ(θ; data) by minimizing −ℓ with the
framework’s L-BFGS optimizer and reports the fit as an MleFit. The
generic machinery lives in mle; concrete likelihood families are added as
sibling submodules that reuse it.
Re-exports§
pub use mle::LogLikelihood;pub use mle::MleFit;pub use mle::fit_mle;pub use types::*;
Modules§
- binomial
- Binomial maximum-likelihood numerics, for the
BinomialLikelihood. - categorical
- Categorical (multinomial-per-observation) likelihood, layered onto the
CategoricalLikelihood. - exponential
- Exponential-distribution maximum-likelihood, for the
ExponentialLikelihood. - mle
- Generic maximum-likelihood estimation for the
MaximumLikelihood. - normal
- Normal (Gaussian) maximum-likelihood numerics, for the
NormalLikelihood. - poisson
- Poisson likelihood: the one-parameter count model for the
PoissonLikelihood. - types
- Plain-data parameter structs for the library’s statistical constructs.