1 of 17

1st ML_INFN Hackathon

Entry Level

​

Statistics and Feedback

Monday, June 21

1

June 2021

ml_infn

Hackathon Statistics

2 of 17

Small report on the hackathon Jun 7th-9th

  • In general, we survived, better than expected
  • The biggest fear was a “technical meltdown” of the infrastructure ...
    • people failing to login, failing instantiation of containers, ...
    • low performance, delays, RAM exhaustion, conflicts of CPU allocation, ...
    • Zoom problems, Zammad Problems, ...
    • late minute changes in files, desync between group areas and GitHub, ...
  • … they simply did not happen
    • virtually 0 problems with IAM (when happening, it was always “network problems on use side”)
    • at the highest utilization, 70+ containers running on INFN Cloud
    • Zoom breakout rooms perfect, Communication with Zammad always working and used also for some time after the end
    • no file lost, overwritten; all well in sync with GitHub

Good!

2

June 2021

ml_infn

Hackathon Statistics

3 of 17

Applications and feedback

3

A large number of application was received very quickly as opening the registration procedure.

Few last-minute withdrawals, usually motivated by important concurrent activities.

in 3 cases we found substitutes, in one other we failed (but we tried...)

June 2021

ml_infn

Hackathon Statistics

4 of 17

Attendance

Peak attendance was registered on Monday morning with 64 people connected to zoom (organizers and tutors excluded).

People not selected for the hackathon connected to the Monday lectures -- ok, it was expected and did not create problems

Some of the registered users did not connect on Tuesday morning -- at least two had a vaccination :)

4

June 2021

ml_infn

Hackathon Statistics

5 of 17

Questions and discussions

​

Most questions asked at the end of each talk through the chat of the zoom application.

Telegram + Zammad used mainly on Tuesday afternoon during the “free study” time.

5

Questions

Replies

Tests

June 2021

ml_infn

Hackathon Statistics

6 of 17

The survey

  • while experience was fresh, we decided to push for a survey, for a few reasons
    • in the plans, we should repeat the hackathon in the second part of the year
    • the hackathon is a new beast, a fully remote hackathon even more

​

Link to the survey (but do not fill it unless you were a student!)

​

https://agenda.infn.it/event/25855/surveys/175

​

Up to now, 20 answers from the ~50 participants … they are slowly coming but already a bit in saturation

6

June 2021

ml_infn

Hackathon Statistics

7 of 17

Feedback from the survey

7

June 2021

ml_infn

Hackathon Statistics

8 of 17

Proficiency before the course

Python

Machine Learning

Cloud Technologies

8

Low

High

Average

Low

High

Average

Low

High

Average

​

​

Useful? 100% says Yes.

But,

  • A bit too general
  • A pre-course one day before the hackathon may be considered
  • A bit too difficult to learn a language by one-self

Comments on the precourse self-training

June 2021

ml_infn

Hackathon Statistics

9 of 17

ML Introductory Lessons (Monday Morning)

Comments:

  • Speakers must try to use the time they have to talk about what is essential, without rushing to stay within the time allowed.
  • Since I liked very much the course, I would like to had more Lessons and more exercises.

​

​

9

June 2021

ml_infn

Hackathon Statistics

10 of 17

Seminars dedicated to state-of-the-art applications within INFN research fields

​

​

Other topics that should have been discussed?

  • Time series
  • Phase Space integration
  • Counting vacua in theoretical physics

10

June 2021

ml_infn

Hackathon Statistics

11 of 17

Hands-on sessions

​

​

Comments:

  • make it easier to save one’s own work
  • some steps were too easy, some too difficult because excessive Python proficiency was assumed

11

On the difficulty level:

Ok

Too easy

Too difficult

June 2021

ml_infn

Hackathon Statistics

12 of 17

The hackathon

12

On the difficulty level:

Ok

Too easy

Too difficult

June 2021

ml_infn

Hackathon Statistics

13 of 17

On Wed afternoon

Comments on the hackathon

  • maybe 4/5 students + 1 tutor is the best
  • I would have preferred more time for the hackathon
  • Only three members in our group actively participated in the hackathon.

13

June 2021

ml_infn

Hackathon Statistics

14 of 17

On the time allocated for the hackathon

14

June 2021

ml_infn

Hackathon Statistics

15 of 17

On the technical setup

15

Will you use INFN Cloud in the future?

No

No as I prefer other clouds

Only with enough resources allocated

Yes

A huge thank you to INFN Cloud for the great support before and during the hackathon!

June 2021

ml_infn

Hackathon Statistics

16 of 17

Take-home messages for the next time

  1. The format works and is appreciated, only fine adjustments should be considered.
  2. Introductory lectures were perceived as a bit too long and/or rushed
    1. the point is “are three days enough for all this?”
  3. Workshop-style talks were appreciated and there was demand for more time and more subject covered (e.g. Theoretical Physics)
  4. Technically, we should prepare and distribute a script to zip everything in the student’s folder in order to make it easier to download one’s own work

​

Should we consider a fourth day?

16

June 2021

ml_infn

Hackathon Statistics

17 of 17

Conclusion

17

The feedback was very generous!

​

Thanks to all of you for having made it possible.

June 2021

ml_infn

Hackathon Statistics