Versions Compared

Old Version 6

changes.mady.by.user Sridhar Rao

Saved on

compared with

New Version 7

changes.mady.by.user Sridhar Rao

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

CategoryDetails
IdITU-ML5G-PS-SODGANS
TitleSynthetic Observability Data Generation using GANs
Description

Observability data can be any of the following:

  1. Telco-Cloud Infrastructure Metrics (servers)
    1. H/W Level
    2. OS Level
    3. Virtualization Level
  2. Metrics/Statistics from Physical Network Elements.
  3. Application (virtualized Network functions) Metrics
  4. System logs (Servers, Applications, etc.)
  5. Metrics/Statistics from centralized orchestration systems (VIM, SDN-Controller, VNFM, NFVO, etc.)
  6. Metrics/Statistics from other control-plane services

Availability of these data for AI/ML researchers, who are not part of the Telco, is very difficult. To solve this availability issue, one approach is to generate synthetic observability data. In this project, we propose to generate this synthetic observability data using GANs.  For this first round only Telco-Cloud Infrastructure Metrics will be considered.

Challenge Track/Theme??? 
Evaluation criteria
  1. Submission of the GANs implementation as Python (or any other language) program.
  2. Least error from the descriminatordiscriminator.
Data sourceReal-World observability data (Telco-Cloud Infrastructure Metrics) will be provided
ResourcesComputing resources can be provided to those who do not have access to one.
Any controls or
restrictions		Anything around data, use of the models, ...
Specification/Paper
reference
Contact

Sridhar K. N. Rao

sridharkn@u.nus.edu


Next Steps:

  1. Submit the above filled template to Thomas (ITU).
  2. ITU will make it publicly available on the registration platform,
  3. Introduction Talk - by Sridhar. Tentative date: 22-06-22. (25 or 27)
  4. Submit the abstract of the talk (including Bio, Pic), 2 weeks before.
  5. Dataset - ITU Platform (if its up to 1 or 2gb) or Github.