Re: [eigen] Eigen for n-dimensional matrices/array

[Rasmus Larsen <rmlarsen@xxxxxxxxxx>]

Eigen does have support for n-dimensional arrays ("tensors") in the unsupported/Eigen/CXX11/src/Tensors tree.

https://bitbucket.org/eigen/eigen/src/7065010756070abd62f1524a5fed29bdbac36b39/unsupported/Eigen/CXX11/src/Tensor/?at=default

Benoit Steiner is the main developer and this code forms the foundation or much of Google's TensorFlow system.

Best,

   Rasmus

On Thu, Oct 6, 2016 at 9:07 AM, Michael Bieri <mibieri@xxxxxxxxx> wrote:

Hi there,

It's still cumbersome to work with more-than-1-dimensional (numeric) arrays in C++. Convenient allocation or indexing is not possible in C++ itself. Not to mention things like slicing...

Eigen is a very helpful library to make it easier. I just wonder: Why does Eigen only support 2 D arrays/matrices? Even in linear algebra, for example in Matlab or NumPy one sometimes use things like a "stack of matrices" or a 3 D image.

Is it somehow possible/planned to handle n-D matrices/arrays?

Best regards,

Michael