[GSoC 2019] Possible Mentorship for a project on implementing uBLAS tensor product and decomposition algorithms

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

[GSoC 2019] Possible Mentorship for a project on implementing uBLAS tensor product and decomposition algorithms

Boost - uBLAS mailing list
Hello, 

My name is Thomas Yang, a 4th year BS/MS student studying Computer Science and Electrical Engineering at Northwestern University. I have used boost in the past for a summer internship, and I would love to learn more about the possibility of completing a GSoC project with the organization, and about contributing to boost numeric libraries in general. 

I noticed that currently in the tensor library, there is only support for tensor arithmetic involving standard tensor n-mode multiplication. Other products which are highly useful in working with large tensors are missing, including the various products mentioned in the project page (e.g. Kronecker product) have not yet been implemented. Furthermore, there is no representation for a tensor decomposition for default dense tensors. I was hoping to pursue a project where I would implement these in the algorithms.hpp header alongside the other products, as well as possibly creating a new tensor decomposition class. This would enable high-order tensors to be represented as products of lesser order tensors, thus enabling users to work with large datasets with uBLAS tensors.

Here is my programming competency test: https://github.com/thomasyang1207/BoostProgrammingAssessment. In my repository, I have also included other open source contributions, as well as personal projects of my own. 

I understand that this email is quite late, and that mentorship opportunities are quite low at this point, but I would love any feedback on my proposal, and if possible, some guidance on possibly doing a GSoC project next year. If mentorship is possible this year, I can quickly complete a proposal within this week. 

Thank you very much, 

Thomas Yang

_______________________________________________
ublas mailing list
[hidden email]
https://lists.boost.org/mailman/listinfo.cgi/ublas
Sent to: [hidden email]
Reply | Threaded
Open this post in threaded view
|

Re: [GSoC 2019] Possible Mentorship for a project on implementing uBLAS tensor product and decomposition algorithms

Boost - uBLAS mailing list
Hello Thomas,

thanks for considering Boost/uBLAS.

The current tensor implementation does not provide any alternative tensor representation such as the tucker, cp or the tensor-train format. We want to definitely support such representations and according operations. However, this year the focus is more on the dense structures. On the long run it would be good for Boost/uBlas to support the above mentioned types and operations.


Your matrix expressions do not include expression templates. Are you going to add this feature?
If you have resources and time left, please write your proposal (google documents) and share the link of your proposal directly with me.

Best,
Cem




Am Di., 26. März 2019 um 04:38 Uhr schrieb Thomas Yang via ublas <[hidden email]>:
Hello, 

My name is Thomas Yang, a 4th year BS/MS student studying Computer Science and Electrical Engineering at Northwestern University. I have used boost in the past for a summer internship, and I would love to learn more about the possibility of completing a GSoC project with the organization, and about contributing to boost numeric libraries in general. 

I noticed that currently in the tensor library, there is only support for tensor arithmetic involving standard tensor n-mode multiplication. Other products which are highly useful in working with large tensors are missing, including the various products mentioned in the project page (e.g. Kronecker product) have not yet been implemented. Furthermore, there is no representation for a tensor decomposition for default dense tensors. I was hoping to pursue a project where I would implement these in the algorithms.hpp header alongside the other products, as well as possibly creating a new tensor decomposition class. This would enable high-order tensors to be represented as products of lesser order tensors, thus enabling users to work with large datasets with uBLAS tensors.

Here is my programming competency test: https://github.com/thomasyang1207/BoostProgrammingAssessment. In my repository, I have also included other open source contributions, as well as personal projects of my own. 

I understand that this email is quite late, and that mentorship opportunities are quite low at this point, but I would love any feedback on my proposal, and if possible, some guidance on possibly doing a GSoC project next year. If mentorship is possible this year, I can quickly complete a proposal within this week. 

Thank you very much, 

Thomas Yang
_______________________________________________
ublas mailing list
[hidden email]
https://lists.boost.org/mailman/listinfo.cgi/ublas
Sent to: [hidden email]

_______________________________________________
ublas mailing list
[hidden email]
https://lists.boost.org/mailman/listinfo.cgi/ublas
Sent to: [hidden email]