bonmi's title

bonmi(W, r, weights = NULL, rsvd.use = FALSE)

Arguments

W: a list of PPMI matrices, with rownames and colnames being the features
r: rank
weights: The weight vector for the PPMI matrices. Default NULL, the weights will be estimated from data.
rsvd.use: Bool. Default FALSE. If TRUE, we will use the 'rsvd' function to calculate the svd, which is much faster than the 'svd' function when r is small.

Value

An embedding matrix.

Examples

set.seed(1)
N = 3000
r = 10
m = 5
p = 0.1
X0 = matrix(rnorm(N*r),nrow=N)
W0 = X0 %*% t(X0)
rownames(W0) = colnames(W0) = paste0('code',1:N)

W = list()
for(s in 1:m){
  ids = which(runif(N)<p)
  Ns = length(ids)
  Es = matrix(rnorm(Ns*Ns,sd=s*0.01),nrow=Ns)
  Es = Es + t(Es)
  Ws = W0[ids,ids] + Es
  W[[s]] = Ws
}

Xhat <- bonmi(W,r,weights=NULL,rsvd.use=FALSE)
#Xhat <- bonmi(W,r,weights=NULL,rsvd.use=TRUE)
codes = rownames(Xhat)
What = Xhat%*%t(Xhat)

id = match(codes,rownames(W0))
Wstar = W0[id,id]

#bonmi's result
norm(What-Wstar,'F')/norm(Wstar,'F')
#> [1] 0.003746704