Project review preparation, Sun 2018-10-28

This is the old pad for Sunday. Please use instead the Pad for Monday.

KPIs

Review the KPI from our 2017 work plan revisions.

KPIs (= key performance indicators) are some measurable ways to evaluate our work.

At the M18 review we presented the current state of our KPIs. It needed improving! Need more data collection…

Aim 1: improve the productivity of researchers in pure mathematics

We need more blog posts of stories.

Aim 2: success stories about odk based VRE deployments

See the list at https://github.com/OpenDreamKit/OpenDreamKit/issues/174

Sunday Rehearsals

Case study: Micromagnetics (Beg)

Start: 14:58 Objective: 15:11 Dissemination: 15:19

Reaction of the public

Demonstrator of VRE and ODK technology for Science

WP3 Component architecture (Luca)

Why do we care about flexibility, portability, …? How do this relate to ODK’s aims? Connect this to ODK’s toolkit approach.

Task list a bit early in the presentation? Deliverables: highlight how does it tie with the rest of ODK

Why did we shift away from cocalc?

Engage users to install software

Engage users to contribute

Sustainability

Sage / SageMath consistency

Strange to mention J. Rueth by name in presentation but nobody else

WP5 HPC (Clément)

Mention MPIR

~~In components mention PPL or/and Normaliz (polytope libraries) ~~ ~~Starting with uses cases milestones (e.g. pictures of a bunch of math objects, … talk to Nicolas) and emphasize

-> Possibilities

Question: how to mention the recommendation without looking defensive? (e.g. no need to repeat it several times)

Why the focus on linbox and combinatorics? What challenges / barrier did we attack? What have we learned?

-> LinBox = linear algebra is central combinatorics = good use case for delicate parallelization

Integer vectors and polytopes: what’s the relation with HPC?

Mention what is SIMD/Cython/Pythran/Cilk

Pythran was just improved by ODK Pythran hard to integrate: we promised we would try to tackle this challenge

When using Cython properly, it shouldn’t be slower than Pythran. Both generate C/C++ code, so the speed should be the same (Jeroen).

-> See benchmarks from https://github.com/videlec/perm_benchmarks

No need to speak about the deliverable merge; focus on content

HPC-GAP: highlight the challenge: HPC-GAP was the main outcome of a previous european project, and had since been dying for years

GAP: maybe let Alex present it? combinatorics (D5.11) + PARI/GP: let Vincent present

~~Many typos: ~~- T5.4, D5.13: Paralell -> Parallel ~~- (slide 25) optimisaiton -> optimisation

Prononciation

WP4 User Interfaces (Vidar)

✓ Start with: when designing the project, we made a bet on technology Jupyter.

✓ That was a huge bet on what technology to base our VRE kit on.

✓ Jupyter received 2017 ACM software system award!

https://blog.jupyter.org/jupyter-receives-the-acm-software-system-award-d433b0dfe3a2

✓ Note that previous recipients of this award included:

https://en.wikipedia.org/wiki/ACM_Software_System_Award

✓ Millions of Jupyter notebooks (see e.g. https://arxiv.org/abs/1810.08055: “as of September 2018, there were more than 2.8 million Jupyter Notebooks shared publicly on GitHub”)

✓ Background on notebook interface: show a rich widget

✓ Three areas of work

Nice screenshots and drawings

✓ Number of kernels: mention the number of kernels contributed by ODK ✓ Kernels ODK contributed to: cling (C++), GAP, Singular, PARI/GP, SageMath, MMT

The notebooks should be on http://opendreamkit.org/try/try (broken link? doesn’t work for me –Katja; fixed now –Vidar) And add the URL on the slides if they are available (or make the pictures clickable)

✓ Typo: “Has met with …”

Active document:

✓ JupyterHub deployment: we contributed to the deployement of many …

Make presentation look better

If possible, Luca would like to do the WP3 talk after this one.

Discuss with Logilab people to report on Jupyter-Simulagora convergence.

Slide additions by Jeroen Demeyer

WP6 Data Knowledge Software Bases (Michael)

Each new slide stays blurred for 1/2 sec. (Maybe particular to Skim under macOS. Not a major issue.)

Typo in slides footer: OKD -> ODK (in “Second OKD review, Oct. ‘18”)

Intereseted -> Interested (slide 9?)

On purpose that the arguments are G,A,p ?

John install a GAP method by calling -> installs

Virtual theories slide: maybe a little hard to read black on purple? (use a lighter shade of purple?)

“MathHub An Portal”: what’s “An” here?

Apocalipse -> Apocalypse

compromized -> compromised docker -> Docker React.SJ -> React.js

“Success story” blog post for Singular mitm contribution?

Decompose one-liner

MitM.Singular(MitM.Gap.orbit(G,A,p)).Ideal().Groebner().sage

into several lines:

o = MitM.Gap.orbit(G,A,p) # the orbit i = MitM.Singular(o).Ideal() # the ideal g = i.Groebner().sage() # the Gröbner basis

Question: getting rid of the A in the orbit calculation, considering that acting on variables is the natural default action for a group on a polynomial?

Suggested puns:

Conclusion slide is quite full: make it lighter? Some slides very busy with text.

In “results of WP6 workshops” slide, could highlight chronology by putting date before location as follows:

WP2 Dissemination (Viviane)

Do not say “we tweet 2-3 times per month”. Say “a lot of our tweets are retweets of community members and ODK components, many of which are also quite active on twitter”

May want to show a page with screenshots of some of most popular tweets as reported by twitter analytics

Remember to mention interaction with the Carpentries: we have several instructors and trainers in the project and even in this room

Monday Rehearsals

Introduction (Nicolas)

Brief explanation of the project, based on the web site Key ideas:

What has happened since the last review:

Case study: Micromagnetics (Beg)

Science and Open Source: what do we learn from each other? Viviane Pons

Demos / posters

WP3 Component architecture (Luca)

WP5 HPC (Clément)

WP4 User Interfaces (Vidar)

Lunch + demos

WP1 Management (Nicolas)

WP6 Data Knowledge Software Bases (Michael)

WP2 Dissemination (Viviane)

Wrapup (Nicolas)

<