match.factors: Align and reorder factors

In zdebruine/LSMF: Least squares matrix factorization

Description

Aligns columns in m.toalign to the ordering of columns in m.alignto based on cosine similarity using bipartite graph matching such that the pairing gives a minimum possible total diagonal cosine distance.

Usage

match.factors(m.alignto, m.toalign)

Arguments

m.alignto	template matrix
m.toalign	matrix to match to template, reorder columns, and return

Details

The Hungarian algorithm is used for bipartite graph matching. The RcppHungarian implementation is natively used. Cosine distance used is from lsmf::sparse.cos with Euclidean norm.

Value

list of components:

• "m.aligned" with m.toalign columns reordered to match m1m2

• "cos" distance of each factor to m1

• "rel.cos" distance as the best distance over the best distance not used in the pairing

• "cost" of the matching

Examples

## Not run: 
data(moca7k)
library(Matrix)
s <- as(sparse.cos(moca7k[,1:1000]), "dgCMatrix")
W1 <- lsmf(s, 10, symmetric = TRUE)$W
W2 <- lsmf(s, 10, symmetric = TRUE)$W

# factors not aligned
plot(as.matrix(W1), as.matrix(W2))
mean(diag(cor(W1,W2)))
# [1] 0.1210305

# factors aligned
W2 <- match.factors(W1, W2)$m.aligned
plot(as.matrix(W1), as.matrix(W2))
mean(diag(cor(W1,W2)))
# [1] 0.9166595


## End(Not run)

zdebruine/LSMF documentation built on Jan. 1, 2021, 1:50 p.m.