# ttclass/ttort.m

Performs TT-orthogonalisation for a tensor train (or for each tensor train in a buffered sum).

## Syntax

    [tt,lognrm]=ttort(tt,direct)


## Arguments

   direct=+1 - {default} gives you left-to-right orthogonality,
direct=-1 - gives right-to-left orthogonality

tt     - tensor train object, possibly with buffered sums


## Outputs

   tt     - tensor train object with all terms in the buffe-
red sum has all of them orthogonalised in the
direction requested

lognrm - if this output is present, all buffered trains
are also normalized, and natural logs of their
norms returned in the vector lognrm. Use this
option if the tensor norm is likely to exceed
realmax()=1.7977e+308.


## Notes

Normally you should not call this subroutine directly.