grow {FamilyRank}

Grow Families

Description

Call to the C++ function that grows the families.

Usage

grow(n, f, d, graph, scores, feat_mat, score_mat, tol, weight_mat, selected)

Arguments

n

Number of features to rank.

f

Number of families to grow.

d

Damping factor

graph

A matrix or data frame representation of a graph object.

scores

A numeric vector of empirical feature scores.

feat_mat

Matrix to store selected features.

score_mat

Matrix to store weighted scores of selected features.

tol

Tolerance

weight_mat

A matrix to store the cumulative weighted scores of selected futures across all families.

selected

Vector indicating whether a feature has been selected yet.

Details

This is the workhorse function for the Family Rank algorithm.

Value

Returns a matrix with 1+2xn.families columns and n.rank rows. The first column is the cumulative feature score for each of the ranked features 1:n.rank. The row number corresponds to the re-indexed feature index. The next n.families columns contain the indices of selected features for each iteration of feature selection. The last n.families columns contain the weigthed scores of selected features for each iteration.

Author(s)

Michelle Saul

References

ADD REFERENCE

Examples

# Toy Example
scores <- c(.6, .2, .9)
graph <- cbind(c(1,1), c(2,3), c(.4, .8))

# initialize matrices 
n <- n.families <- length(scores)
feat.mat <- score.mat <- matrix(0, nrow = n, ncol = n.families)
feat.mat[1,] <- order(scores, decreasing = TRUE)
score.mat[1,] <- sort(scores, decreasing = TRUE)

# Grow families
mats <- grow(n = n, f = n.families, d = 0.5, graph = as.matrix(graph), 
             scores = scores, 
             feat_mat = feat.mat, score_mat = score.mat, tol = 0, 
             weight_mat = as.matrix(scores), selected = rep(1, n))
# Selected Feature Matrix
## columns represent familes
## rows represent iterations
## values indicate indices of selected features
feat.mat <- mats[, 2:(n.families+1)]
feat.mat
# Corresponding Score Matrix
## columns represent familes
## rows represent iterations
## values indicate max weighted score of selected features
score.mat <- mats[, (n.families+2):(1+2*n.families)]
score.mat