BBKNN(Batch-balancing kNN)-adjusted SVD

BBKNN(Batch-balancing kNN)-adjusted SVD

rcpp_mmutil_bbknn_mtx(
  mtx_file,
  r_batches,
  knn,
  RANK,
  RECIPROCAL_MATCH = TRUE,
  TAKE_LN = TRUE,
  TAU = 1,
  COL_NORM = 10000,
  EM_ITER = 0L,
  EM_TOL = 1e-04,
  KNN_BILINK = 10L,
  KNN_NNLIST = 10L,
  LU_ITER = 5L,
  row_weight_file = "",
  NUM_THREADS = 1L,
  BLOCK_SIZE = 10000L,
  USE_SINGULAR_VALUES = FALSE
)

Arguments

mtx_file

data file (feature x n)

r_batches

batch names (n x 1)

knn

kNN parameter k

RANK

SVD rank

RECIPROCAL_MATCH

do reciprocal match (default: TRUE)

TAKE_LN

take log(1 + x) trans or not

TAU

regularization parameter (default: 1)

COL_NORM

column normalization

EM_ITER

EM iteration for factorization (default: 0)

EM_TOL

EM convergence (default: 1e-4)

KNN_BILINK

num. of bidirectional links (default: 10)

KNN_NNLIST

num. of nearest neighbor lists (default: 10)

LU_ITER

LU iteration

row_weight_file

row-wise weight file

NUM_THREADS

number of threads for multi-core processing

BLOCK_SIZE

number of columns per block

USE_SINGULAR_VALUES

Weight factors by the corresponding SVs

Value

a list of (1) factors.adjusted (2) D (3) V (4) knn