Re: [boost] [gsoc18] Mentor Request for uBLAS (prefer support for tensors)

Hi Cem,

I'm sorry for such a reply. What you do is indeed really interesting. I had
a PhD student 3 years ago for a GSOC and the result was outstanding.
I'm really interested in your project.

Do you think you can submit a proposal for this year ?
Proposals are very open and it's not unusual we take a student with
something different that we "officially" proposed. So I'm happy to see
something which would fit your phd work too.

Cheers,
David

On Thu, Mar 15, 2018 at 12:56 PM, Cem Bassoy via ublas <
[email protected]> wrote:

> Hello,
>
> my name is Cem. I am a PhD candidate working at the Fraunhofer Institute
> of Optronics, System Technologies and Image Exploitation.
>
> I am a passionate programmer in C++ since I am student that I intensively
> use on a daily basis. My PhD thesis is about implementing fast tensor
> operations on shared memory systems using C++ (C++14) and OpenCL for GPUs.
> By tensor operations I mean those that are relevant in the realm of
> numerical multilinear algebra. I have already implemented tensor
> contractions and elementwise tensor operations for tensors and tensor views
> with arbitrary number of dimensions where the rank can be compile-time or
> runtime variable. For testing purposes I have implemented tensor data
> structures that allow to set the rank, dimensions and the layout at runtime
> (They could also be changed to be compile-time parameters).
>
> I also want to implement common tensor algorithms such as the higher-order
> singular value decomposition, orthogonal iterator in order to support new
> algorithms for different types of applications. Using my library I have
> generalized and implemented the biconjugate gradient and Jacobi methods for
> multilinear systems that are described in
> "Solving Multilinear Systems via Tensor Inversion
> M. Brazell, N. Li, C. Navasca, and C. Tamon
> SIAM Journal on Matrix Analysis and Applications201334:2, 542-570"
>
> I presented the implementation and design of tensor template functions
> using iterators in MeetingC++ in Berlin 2017:
> https://youtu.be/8P-sDH9XJAc
>
> If your are interested in the details of my current implementation, please
> have a look at https://arxiv.org/abs/1711.10912
>
> I would love to contribute to uBlas and to adjust the data structures
> according to the needs of the Boost community.
>
> Best regards,
>
> Cem
>
> _______________________________________________
> ublas mailing list
> [email protected]
> https://lists.boost.org/mailman/listinfo.cgi/ublas
> Sent to: [email protected]
>
>