federated fashion

Description Usage Arguments Value

This function creates the files and file system required to train a linear model in a distributed fashion.

initializeDistributedModel(formula, model = "LinearModel",
  optimizer = "gradientDescent", out_dir = getwd(), files, epochs,
  learning_rate, mse_eps, save_all = FALSE, file_reader,
  overwrite = FALSE)

`formula`	[`formula`] Formula analog to the formula call in `lm`.
`model`	[`character(1)`] Character indicating the model we want to use.
`optimizer`	[`character(1)`] Character indicating which optimizer we want to use.
`out_dir`	[`character(1)`] Direction for the output files.
`files`	[`character`] Vector of file destinations. Each element must point to one dataset.
`epochs`	[`integer(1)`] Number of maximal iterations. Could be less if the "epsilon criteria" is hit.
`learning_rate`	[`numeric(1)`] The step size used for gradient descent. Note: If the mse is not improving the step size is shrinked by 20 percent.
`mse_eps`	[`numeric(1)`] Relativ improvement of the MSE. If this boundary is undershot, then the algorithm stops.
`save_all`	[`logical(1)`] If set to TRUE, all updates are stored within the out_dir.
`file_reader`	[`function`] Function to read the datasets specified in files.
`overwrite`	[`logical(1)`] Flag to specify whether to overwrite an existing registry and model or not.