Compute Task-Based Parcel Similarity Graph (W_task) from Encoding Weights
Source:R/task_graph_construction.R
compute_W_task_from_encoding.RdCalculates a sparse, z-scored similarity graph between parcels based on their encoding weight profiles for a set of features.
Usage
compute_W_task_from_encoding(
encoding_weights_matrix,
parcel_names,
k_conn_task_pos,
k_conn_task_neg,
similarity_method = "pearson"
)Arguments
- encoding_weights_matrix
A numeric matrix (`V_p x N_features`) where `V_p` is the number of parcels and `N_features` is the number of encoding features. Each row represents the encoding weight profile for a parcel.
- parcel_names
A character vector of length `V_p` specifying parcel names.
- k_conn_task_pos
Non-negative integer. Number of strongest positive connections to retain per parcel during sparsification.
- k_conn_task_neg
Non-negative integer. Number of strongest negative connections to retain per parcel during sparsification.
- similarity_method
Character string or function. Specifies the method to compute the initial `V_p x V_p` similarity matrix. If "pearson" (default) or "spearman", `stats::cor` is used on the transposed input (to compare rows/parcels). If a function, it must take `encoding_weights_matrix` (V_p x N_features) as input and return a `V_p x V_p` numeric matrix.