t3_operations

Structural operations on ragged t3 data.

Squash tails, segment/concatenate, stack/unstack, broadcast-to-common-stack, core-shape utilities, and the weighted-layer core ops (t3_absorb_weights / t3_{concatenate,kronecker}_weights / t3_weights_consistent).

Functions

t3_broadcast_to_common_stack(tucker_cores, tt_cores)	Broadcast every core of a T3 up to the common (broadcast) stack of all its cores.
t3_absorb_tucker_into_tt(tucker_cores, tt_cores)	Absorb each Tucker core into its TT core, replacing the mode (n) leg with the physical (N)
t3_squash_tails(data)	Collapse the leading/trailing TT bonds of a T3 to one, without changing the represented tensor.
t3_segment(data[, start, stop])	Contiguous mode-segment of a T3: slice both core families over modes start:stop.
t3_concatenate(xx)	Concatenate a sequence of T3 segments into one T3 (inverse of t3_segment()).
tucker_change_core_shapes(tucker_cores, new_shape, ...)	Increase/decrease Tucker and/or TT ranks for TT cores using zero padding/truncation.
t3_stack(xx)
t3_unstack(x)	Given multiple stacked T3s, this unstacks them
t3_core_shapes(shape, tucker_ranks, tt_ranks[, ...])	Determines the shapes of the T3 cores based on the ranks.
t3_sum(x[, axis])	Sum over axes of TuckerTensorTrain.
t3_absorb_weights(x0, weights)	Contract diagonal edge weights into a Tucker tensor train's cores (shape-preserving).
t3_weights_consistent(x0, weights)	True iff the weight tuple's lengths, per-edge ranks, and stack_shape match the T3 x0
t3_concatenate_weights(weights_A, weights_B)	Per-edge concatenation of two weight tuples (the + / direct-sum combine: ranks add). Last-axis
t3_kronecker_weights(weights_A, weights_B)	Per-edge Kronecker product of two weight tuples (the Hadamard * combine: ranks multiply). The