tuning: Tunes the methods to have the best set of parameters
In agapow/subtypr: Robust and Validated Patient Subtyping for Precision Medicine

Description Usage Arguments Details Value

tuning considers a list of values of parameters (grid_support) to be tested (all the combinations are tested) and find the set of parameters that have the best value for the metric selected.

1
2
3

tuning(data_list, method, grid_support, metric, true_partition = NULL,
  parallel = TRUE, ncores = NULL, save_results = FALSE,
  file_name = "./tuning_result.RData", plot = TRUE, verbose = TRUE)

`data_list`	a list of data matrices with continuous data of format samples x features (with the same number of samples).
`method`	a string being the name of the built-in method to be used or a 'method list' with the same format as the built-in ones. See Details
`grid_support`	a list with a set of value for each parameter to be tuned. The typo has to be correct. Use formals(method$func) to know all the available parameters.
`metric`	a string being the name of the built-in metric to be used or a 'metric list' with the same format as the built-in ones. If the chosen metric is internal, be sure to choose a metric corresponding to the kind of data returned by the chosen method.
`true_partition`	a factor or integer vector being a classification of the patients. It will be used with external metrics.
`parallel`	logical, TRUE for parallel computing (default and recommanded).
`ncores`	integer, the number of core to be used for the parallel computation. If NULL, it takes the maximum number of cores - 1, using the function `detectCores`.
`save_results`	logical, to save the result of tuning in `path_to_file`
`file_name`	filename finishing by .RData or .rda
`plot`	logical, set TRUE to plot the metric evaluation of each grid point to have an overview of the impact of the variation of the parameters on the value of the considered metric.
`verbose`	logical, set FALSE to avoid printing of informations
`path_to_file`	if save_results, results of the tuning are save to this path

To use a custom method, just pass your function method. Your function has to respect the following structure:

Input: function(data_list, param_1, param_2, ..... , param_x, ...)
Output: A list with:
- partition: an integer vector labelling the sample
- element_for_metric: a character that indicates the name of the element used with an internal metric for validation. It can be NULL.
- output_1
- ...
- output_x

a list:

$metric_val contain the best value of the metric among all the grid
$parameters is the corresponding set of parameters
$method_used contains the name of the tuned method, to know where the result comes from.
$method_res contains the result of the method, i.e. a partition and the data returned by the method, to allow external and/or internal validation by other metrics.