Skip to content
Lea Goetz edited this page Jan 23, 2018 · 26 revisions

Information and schedule about the 2017 Shogun hackathon in Budapest. Kindly sponsored by NumFOCUS.

Check out the blog about the hackathon!

Roadmap 2018 discussion

People named behind a topic will create a wiki page that outlines the intermediate steps, timing, necessary resources, or anything else that is needed to make it happen.

For users

  • We want to put continuous efforts into making installation of Shogun and running first examples a matter of minutes. (Heiko)

  • We will explicitly develop documentation for users, core-developers, and ML-developers. (Giovanni, Esben)

  • We want to add JavaScript bindings to Shogun's interfaces, allowing Shogun to be used from the NodeJS community. (Viktor)

  • We want to compile Shogun into WebAssembly. (Fernando)

  • We will bridge Shogun with existing ML frameworks, such as XGBoost, Keras, TensorFlow, or even pretrained networks, in order to expose their functionality through our interfaces. (Sergey)

  • We will expose Shogun via other popular ML frameworks, such as bioconductoR, sklearn, and scilab, in order to indirectly increase our user-base. (Gunnar, Lea)

  • We will buy hardware to continue running the Shogun cloud at manageable costs. (Gunnar)

For developers

  • We want to change Shogun's API to a more simple, demangled version, and freeze it subsequently [link to prototype]. (Heiko)
  • Using the new tags framework, we will move away from Shogun's monolythic structure towards a plugin based one. (Heiko, Viktor, Sergey)
  • We will develop a use-case for ML scientists to use Shogun as a vehicle for implementing their algorithms, harnessing ease of installation, numerical computing, and interfaces. (Esben, Michele)
  • We will put continuous effort into modernising the low-level parts of the Shogun library, using latest C++ standards and libraries. (Viktor)

Outreach / Education / Science

  • We will write a JMLR paper that introduces Shogun's latest features, and compares it to popular packages. (Heiko)
  • We will encourage our community to produce a series of blog posts that showcase end to end data-science applications using Shogun. (Lea, Pan)
  • We will apply to GSoC 2018 and continue to develop close relationship with participating students. (The team)
  • We will present Shogun at events in the ML space. (Pan)
  • We will organise two hackathons (April and November-ish) and invite our community members. (Gunnar, Heiko)
  • On our website, we will explicitly approach users and developers. (Lea)

Group sessions

For your topic

  • First: pick person for notes, timekeeping, and presentation
  • Define your topic! What does it mean/involve, what is the current state?
  • What is good? What is the problem?
  • What can be improved?
  • How can it be improved?
  • Action items (github project!)
  • Optional: draft

Friday

  • Everyone arrives.
  • 19:00 Vague dinner plans (either in flat or out), socialising Visiting some of the https://ruinpubs.com/ (if you have any wishes which one to definitely visit please let @Viktor know) - check the picture on google as well

Saturday

  • 0830 Breakfast in flat
  • 0930 Departure to venue
  • 1000 Intro
  • 1030 Group session: website (3rd floor), doc (2nd), devops
  • 1130-1145 Coffee break
  • 1230 Presentation / discussion
  • 1300 Lunch
  • 1430 Group session: modernise (1st floor), governance (2nd floor), issues (3rd floor)
  • 1530-1545 Coffee break
  • 1630 Presentation / discussion
  • 1700 Putting session results into Wiki, iterating on results, buffer time
  • 1800 End of day
  • 2000 Dinner
  • Next round of the famous Budapest Ruin pubs :)

Sunday

  • 0830 Breakfast in flat
  • 0930 Departure to venue
  • 1030 Group session: BSD release (1), Base ML API (2), Blog post about hackathon (3)
  • 1130-1145 Coffee break
  • 1230 Presentation / discussion
  • 1300 Lunch
  • 1430 Group session
  • 1530-1545 Coffee break
  • 1630 Presentation / discussion
  • 1700 Putting session results into Wiki, iterating on results, buffer time
  • 1800 End of day
  • 1930 Spa!

Monday

  • Some of us will stay around, sightseeing, hacking, or chatting. Feel free to hang out.

Group sessions

Topics

Participants chose a number of sessions they would like to participate in

  • Modernising C++ core
  • Shogun's future
  • Dev-ops
  • API specifications
  • Website
  • Governance
  • GSoC 2018
  • Issue sprint
  • Documentation sprint
  • BSD release sprint
  • Blog post about hackathon

The last one was a suggestion "Put word out about shogun/the hackathon (Not least for the NumFOCUS people)". Good point, we actually promised a blog post that when we asked NF for money :) So maybe we can recruit a few more people for writing that.

The group session schedule is computed to be clash free according to everyone’s questionnaire replies. I.e. everyone can attend all the sessions they voted for, see Heiko's schedule notebook and Viktor's PCA extension Correct us if there are any bugs in there :)

Clone this wiki locally