The linear_models module includes:
- OLS linear regression with maximum likelihood parameter estimates via the normal equation.
- Includes optional weight arguments for weighted least squares
- Supports batch and online coefficient updates.
- Ridge regression / Tikhonov regularization with maximum likelihood parameter estimates via the normal equation.
- Logistic regression with maximum likelihood parameter estimates via gradient descent.
- Bayesian linear regression with maximum a posteriori parameter estimates via conjugacy
- Known coefficient prior mean and known error variance
- Known coefficient prior mean and unknown error variance
- Naive Bayes classifier with Gaussian feature likelihoods.
- Generalized linear model with identity, log, and logit link functions.
