Struct TtDecomposition 

pub struct TtDecomposition {
    pub cores: Vec<Tensor>,
}

Tensor-Train (TT) decomposition result.

Represents a tensor as a chain of 3D cores: X(i_1, ..., i_N) = G_1(i_1) · G_2(i_2) · ... · G_N(i_N) where each G_k(i_k) is an r_{k-1} × r_k matrix slice, and core k has shape r_{k-1} × n_k × r_k.

Fields

cores: Vec<Tensor>

TT-cores, each a 3D tensor of shape [r_{k-1}, n_k, r_k].

Implementations

impl TtDecomposition

pub fn ranks(&self) -> Vec<usize>

Returns the TT-ranks [r_0, r_1, ..., r_N] where r_0 = r_N = 1.

pub fn reconstruct(&self) -> SolverResult<Tensor>

Reconstructs the full tensor from TT-cores.

pub fn fit_error(&self, original: &Tensor) -> SolverResult<f64>

Computes the relative fit error: ||X - X_hat|| / ||X||.

pub fn compression_ratio(&self, original_shape: &[usize]) -> f64

Compression ratio: original elements / decomposition elements.

pub fn tt_round(&self, max_rank: usize) -> SolverResult<TtDecomposition>

TT-rounding: truncates TT-ranks to at most max_rank.

Performs left-to-right QR orthogonalization, then right-to-left truncated SVD to reduce ranks.

Trait Implementations

impl Clone for TtDecomposition

fn clone(&self) -> TtDecomposition

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for TtDecomposition

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.