t3_svd#

T3-SVD: minimal-rank reduction / rank truncation of ragged t3 data.

t3svd (the production sweep), t3_rank_adjustment_sweep, and the dense reference implementations (dense_tucker_svd/dense_ttsvd/dense_t3svd) used for verification. Design + minimal-rank discussion: the docs/t3svd_* notes.

Functions#

t3svd(x[, max_tt_ranks, max_tucker_ranks, rtol, atol, ...])

Compute (truncated) T3-SVD of TuckerTensorTrain.

t3_rank_adjustment_sweep(x[, direction])

A single lossless directional sweep that drops structurally-redundant ranks (re-SVD each Tucker

dense_tucker_svd(T[, min_ranks, max_ranks, rtol, atol])

Compute Tucker decomposition and matricization singular values for dense tensor.

dense_ttsvd(T[, min_ranks, max_ranks, rtol, atol])

Compute tensor train (TT) decomposition and unfolding singular values for dense tensor.

dense_t3svd(T[, stack_shape, max_tucker_ranks, ...])

Compute TuckerTensorTrain and edge singular values for dense tensor.