.. _glossary-label: Glossary ======== Abbreviations used in the documentation, the examples and the API reference. .. glossary:: :sorted: AD Automatic differentiation. Derivatives of a computation are obtained by applying the chain rule to its elementary operations. ``torchtt`` builds on the PyTorch autograd engine, so gradients can be taken with respect to the TT cores (see :mod:`torchtt.grad`). AMEn Alternating minimal energy method :cite:p:`dolgov2014alternating`. It updates one TT core at a time and enriches the approximation with a low-rank approximation of the residual, so the TT ranks adapt during the iteration. ``torchtt`` uses it to solve linear systems (:func:`torchtt.solvers.amen_solve`), for matrix products (``method='amen'`` in :func:`torchtt.matvec` and :func:`torchtt.matmat`), for elementwise division (:func:`torchtt.elementwise_divide`) and for cross interpolation (``method='amen'`` in :func:`torchtt.interpolate.function_interpolate`). BiCGSTAB Biconjugate gradient stabilized method, an iterative solver for non-symmetric linear systems. :func:`torchtt.solvers.amen_solve` can use it for its local systems (``local_solver=2``). DMRG Density matrix renormalization group. The name comes from quantum physics :cite:p:`white1992density`; for the TT format it denotes schemes that optimize two neighbouring TT cores at a time, which lets the TT ranks adapt :cite:p:`oseledets2011dmrg,savostyanov2011fast`. ``torchtt`` uses it for matrix-vector products (``method='dmrg'`` in :func:`torchtt.matvec`), elementwise products (``method='dmrg'`` in :func:`torchtt.hadamard`) and cross interpolation (``method='dmrg'`` in :func:`torchtt.interpolate.function_interpolate`). GMRES Generalized minimal residual method, an iterative solver for non-symmetric linear systems. :func:`torchtt.solvers.amen_solve` uses it for its local systems by default (``local_solver=1``). QR QR decomposition. It factors a matrix into a matrix with orthonormal columns (:math:`Q`) and an upper triangular matrix (:math:`R`). It is used to orthogonalize TT cores. QTT Quantized tensor train :cite:p:`oseledets2010approximation,khoromskij2011quantics`: the TT format applied to a tensor that has been reshaped so that every mode has size 2. See :ref:`qtt-label`. SVD Singular value decomposition. Truncating it is how TT ranks are reduced, both when a full tensor is decomposed and when a TT tensor is rounded. TT Tensor train :cite:p:`oseledets2011tensor`. A format that represents a :math:`d`-dimensional tensor by :math:`d` three-dimensional tensors, the TT cores. See :ref:`about-tt-label`. TTM TT matrix A linear operator (tensor operator) in the TT format. Its cores are four-dimensional, with one row and one column index each. A ``torchtt.TT`` is a TT matrix when its shape is given as a list of ``(rows, columns)`` tuples.