Add settings matrix rows
Usage
add_settings_matrix_rows(
settings_matrix,
seed = NULL,
nrows = 0,
min_removed_inputs = 0,
max_removed_inputs = sum(startsWith(colnames(settings_matrix), "inc_")) - 1,
dropout_dist = "exponential",
min_alpha = NULL,
max_alpha = NULL,
min_k = NULL,
max_k = NULL,
min_t = NULL,
max_t = NULL,
alpha_values = NULL,
k_values = NULL,
t_values = NULL,
possible_snf_schemes = c(1, 2, 3),
clustering_algorithms = NULL,
continuous_distances = NULL,
discrete_distances = NULL,
ordinal_distances = NULL,
categorical_distances = NULL,
mixed_distances = NULL,
distance_metrics_list = NULL,
snf_input_weights = NULL,
snf_domain_weights = NULL,
retry_limit = 10
)
Arguments
- settings_matrix
The existing settings matrix
- seed
set a seed for the random matrix generation. Setting this value will change the seed of the global environment.
- nrows
Number of rows to generate for the settings matrix.
- min_removed_inputs
The smallest number of input dataframes that may be randomly removed. By default, 0.
- max_removed_inputs
The largest number of input dataframes that may be randomly removed. By default, this is 1 less than all the provided input dataframes in the data_list.
- dropout_dist
Parameter controlling how the random removal of input dataframes should occur. Can be "none" (no input dataframes are randomly removed), "uniform" (uniformly sample between min_removed_inputs and max_removed_inputs to determine number of input dataframes to remove), or "exponential" (pick number of input dataframes to remove by sampling from min_removed_inputs to max_removed_inputs with an exponential distribution; default).
- min_alpha
The minimum value that the alpha hyperparameter can have. Random assigned value of alpha for each row will be obtained by uniformly sampling numbers between
min_alpha
andmax_alpha
at intervals of 0.1. Cannot be used in conjunction with thealpha_values
parameter.- max_alpha
The maximum value that the alpha hyperparameter can have. See
min_alpha
parameter. Cannot be used in conjunction with thealpha_values
parameter.- min_k
The minimum value that the k hyperparameter can have. Random assigned value of k for each row will be obtained by uniformly sampling numbers between
min_k
andmax_k
at intervals of 1. Cannot be used in conjunction with thek_values
parameter.- max_k
The maximum value that the k hyperparameter can have. See
min_k
parameter. Cannot be used in conjunction with thek_values
parameter.- min_t
The minimum value that the t hyperparameter can have. Random assigned value of t for each row will be obtained by uniformly sampling numbers between
min_t
andmax_t
at intervals of 1. Cannot be used in conjunction with thet_values
parameter.- max_t
The maximum value that the t hyperparameter can have. See
min_t
parameter. Cannot be used in conjunction with thet_values
parameter.- alpha_values
A number or numeric vector of a set of possible values that alpha can take on. Value will be obtained by uniformly sampling the vector. Cannot be used in conjunction with the
min_alpha
ormax_alpha
parameters.- k_values
A number or numeric vector of a set of possible values that k can take on. Value will be obtained by uniformly sampling the vector. Cannot be used in conjunction with the
min_k
ormax_k
parameters.- t_values
A number or numeric vector of a set of possible values that t can take on. Value will be obtained by uniformly sampling the vector. Cannot be used in conjunction with the
min_t
ormax_t
parameters.- possible_snf_schemes
A vector containing the possible snf_schemes to uniformly randomly select from. By default, the vector contains all 3 possible schemes: c(1, 2, 3). 1 corresponds to the "individual" scheme, 2 corresponds to the "domain" scheme, and 3 corresponds to the "twostep" scheme.
- clustering_algorithms
A list of clustering algorithms to uniformly randomly pick from when clustering. When not specified, randomly select between spectral clustering using the eigen-gap heuristic and spectral clustering using the rotation cost heuristic. See ?generate_clust_algs_list for more details on running custom clustering algorithms.
- continuous_distances
A vector of continuous distance metrics to use when a custom distance_metrics_list is provided.
- discrete_distances
A vector of categorical distance metrics to use when a custom distance_metrics_list is provided.
- ordinal_distances
A vector of categorical distance metrics to use when a custom distance_metrics_list is provided.
- categorical_distances
A vector of categorical distance metrics to use when a custom distance_metrics_list is provided.
- mixed_distances
A vector of mixed distance metrics to use when a custom distance_metrics_list is provided.
- distance_metrics_list
List containing distance metrics to vary over. See ?generate_distance_metrics_list.
- snf_input_weights
Nested list containing weights for when SNF is used to merge individual input measures (see ?generate_snf_weights)
- snf_domain_weights
Nested list containing weights for when SNF is used to merge domains (see ?generate_snf_weights)
- retry_limit
The maximum number of attempts to generate a novel row. This function does not return matrices with identical rows. As the range of requested possible settings tightens and the number of requested rows increases, the risk of randomly generating a row that already exists increases. If a new random row has matched an existing row
retry_limit
number of times, the function will terminate.