张量分解算法的Julia实现
Available functions
The following functions for Tucker decompositions, except for sshopm, return a Tucker, which contains factors::Vector{Matrix{Float64}}, core::Array{Float64} (1-dimensional array for Kruskal decompositions), and the relative reconstruction error error::Float64.
High-order SVD (HOSVD) [3] hosvd{T,N}(tnsr::StridedArray{T,N}, core_dims::NTuple{N, Int}; pad_zeros::Bool=false, compute_error::Bool=false)
Canonical polyadic decomposition (CANDECOMP/PARAFAC) candecomp{