MultiCX: MultiCX Tensor Decomposition

In ccTensor: CUR/CX Tensor Decomposition

Description

The input data is assumed to be a tensor. MultiCX decomposes the tensor into a core tensor and some factor matrices. The factor matrices are not estimated values but the actual column vectors sampled from the unfolded matrix in each mode.

Usage

MultiCX(Y, rank=NULL, modes=1:3, thr=0.9,
  c.method=c("best.match", "random", "exact.num.random", "top.scores"))

Arguments

Y	The input tensor (e.g. N times M times L).
rank	The number of low-dimension of factor matrices (e.g. J1, J2, and J3). If this argument is not specified or specified as NULL, the low-dimension is estimated based on the cumulative singular value (Default: NULL).
modes	The vector of the modes on whih to perform the decomposition (Default: 1:3 <all modes>).
thr	The threshold to determine the low-dimension of factor matrices. The value must be range 0 to 1 (Default: 0.9).
c.method	The column sampling algorithm (Default: best.match).

Value

U: Core tensor (e.g. J1 times J2 times J3). C: Factor matrices (e.g. C_1: ????????) RecError : The reconstruction error between data tensor and reconstructed tensor from C and X.

Author(s)

Koki Tsuyuzaki

References

Maria F. K. B. et. al. (2019). Multidimensional CX Decomposition of Tensors. WCNPS

Examples

    library("ccTensor")
    library("nnTensor")
    # Test data
    tensdata <- toyModel(model = "CP")
    # Simple usage
    out <- MultiCX(tensdata, rank=c(3,4,5))

ccTensor documentation built on Aug. 12, 2021, 9:07 a.m.