Struct TensorTrain 

pub struct TensorTrain {
    pub cores: Vec<TtCore>,
    pub shape: Vec<usize>,
}

A tensor in Tensor Train (TT / MPS) format.

A d-way tensor T of shape (n_0, …, n_{d-1}) is represented as a product of 3-way cores G_k of shape (r_{k-1}, n_k, r_k) where r_0 = r_d = 1.

Fields

cores: Vec<TtCore>

The TT cores. Core k has shape (r_{k-1}, n_k, r_k). Each core is stored as a flat Vecf64` in C-order.

shape: Vec<usize>

Mode dimensions n_k.

Implementations

impl TensorTrain

pub fn evaluate(&self, idx: &[usize]) -> f64

Evaluate the TT representation at a given multi-index.

Contracts the cores left-to-right: result = G_0[i_0] G_1[i_1] … G_{d-1}[i_{d-1}]. Each G_k[i_k] is the r_{k-1}×r_k slice.

pub fn from_dense( tensor: &DenseTensor, max_rank: usize, tol: f64, ) -> TensorTrain

Convert a full dense tensor to Tensor Train format via left-to-right SVD rounding (TT-SVD algorithm).

max_rank caps each TT rank; tol is relative truncation tolerance.

pub fn frobenius_norm_approx(&self) -> f64

Frobenius norm of the TT tensor (contract all cores).

impl TensorTrain

pub fn frobenius_norm(&self) -> f64

Compute the Frobenius norm via contraction of the full TT.

For small tensors only — reconstructs fully.

pub fn to_dense(&self) -> DenseTensor

Reconstruct to a dense DenseTensor.

pub fn dot_product(&self, other: &TensorTrain) -> f64

Dot product <TT_a, TT_b> between two TensorTrains with the same shape.

Uses the supercore contraction (quadratic in ranks).

pub fn scale(&self, s: f64) -> TensorTrain

Scale all cores by a scalar.

Trait Implementations

impl Clone for TensorTrain

fn clone(&self) -> TensorTrain

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TensorTrain

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

Formats the value using the given formatter.