initialize: Initialize a tundraContainer object.
In syberia/tundra: Standardized container format for deployment of classifiers.

Initialize a tundraContainer object.

1 2	initialize(keyword, train_function = identity, predict_function = identity, munge_procedure = list(), default_args = list(), internal = list())

`keyword`	character. The name of the classifier; for example, "lm" or "knn".
`train_function`	function. The function used to train the model. Its first argument will be a data.frame, and the second argument a list of additional parameters used for training the model.
`predict_function`	function. The function used to predict on new datasets. Its first argument will be a data.frame, the dataset to predict on, and its second (optional) argument will be additional parameters used for prediction output (such as whether to return a probabilistic or absolute value).
`munge_procedure`	list. A list of trained `mungepiece`s to apply to data sets during prediction.
`default_args`	list. A list of default arguments to provide to the second argument of the `train_function`. The additional arguments provided to the `tundraContainer`'s `train` method will be merged on top of these defaults.
`internal`	list. Internal metadata that should accompany the model. Usually this is domain/organization specific, and can include things such as a list of primary keys used for training the model, identifiers or names of data sources used for training the model, etc. It is a playground entirely under your control, and can be used by other packages or a production server hosting the model to achieve additional behavior.