Save a data copy in fitted objects

[jolars]

Consider storing a copy of x and y (feature matrix and response) in the fitted object from sgdnet().

Pros

• Follows conventions from most modelling objects in R.
• Makes it more convenient to call predict methods.

Cons

• Large data objects are may be prohibitive to copy.
• Encourages predictions using "past" data, which could be considered bad form.

Possible solutions

• Use a thin argument in sgdnet() to let the user decide if they want to store the data object.
• Attempt to retrieve the data object from memory (throwing a warning while doing so).
• Use a option, such as sgdnet.store_data so that users can decide if they want to store the data.