Teaching Machine Learning Workshop at ECML-PKDD 2020
https://teaching-ml.github.io/2020/

Roles 09/14/ 2020:
    Welcome and presentation chair: Heidi Seibold
    Technical chair: Oliver Guhr
    Wrap-Up and Farewell: Peter Steinbach


Agenda:
    
09.00 am         Welcome 
09.15 am         Paper Presentations     
10.05 am         Coffee break
10.25 am         Workshop (Discussion in Breakout Rooms)  
12.00 pm         Lunch  
01.00 pm         Wrap-Up Session  
01.30 pm         Farewell and Next Steps   
01.45 pm         End  

Roll call + Check in
(sign-in to stay connected, not required)

• Miriam Elia / miriam@elianet.de / she, her /  @mikaso_e
• Heidi Seibold / heidi.seibold@helmholtz-muenchen.de / she, her / Twitter: @HeidiBaya
• Peter Steinbach / p.steinbach@hzdr.de / he, his / twitter: @psteinb_ 
• Alexander Schiendorfer / alexander.schiendorfer@gmail.com / he, his / twitter: @schienal
• Karsten Lübke / karsten.luebke@fom.de / he/his / twitter: @luebby42
• Daniela Huppenkothen / dhuppenk@uw.edu / she/her / twitter: @Tiana_Athriel
• Javier Garcia-Algarra: javier.algarra@u-tad.com TW @jgalgarra
• Rebecca Fiebrink / r.fiebrink@arts.ac.uk / she/hers / @rebeccafiebrink
• Oliver Guhr / oliver.guhr@htw-dresden.de / he, his / twitter: @oliverguhr
• Samantha Monty / samantha.monty@uni-wuerzburg.de / she/her / @SamanthaMatoush
• Carola Gajek / gajek@isse.de / she, her / twitter: @CGroot21
• Cornelia Gamst / cornelia.gamst@ki-campus.org / she / @gamstc
• Benedikt Weygandt / benedikt.weygandt@ki-campus.org / he/him 
• Katherine M. Kinnaird / kkinnaird@smith.edu / she, her, hers / @kmkinnaird
• Azqa nadeem / azqa.nadeem@tudelft.nl / she,her / @azqa_nadeem

Paper presentations
Paper presentations will be recorded and published on YouTube (if presenters don't object)

On YouTube
“Teaching Computational Machine Learning (without Statistics)” by Katherine M. Kinnaird

• Thank you very much! Maybe I missed it, but maybe you can give more background on the students and how this course is integrated in curriculum?
  • Thank you for your question. This course was offered in a computer science department at Smith College, a small liberal arts college (SLAC) and a historically women's college in the United States. The college has about 2600 undergraduate students (and while there are post-baccalaurate programs at Smith College, there is no such program in the computer science department). This year, 56 students graduated with degrees in computer science. This past year, students in this course had taken fewer than 12 courses in computer science, and all self-identified as women. 
  • In terms of integrating into the curriculum, this course was designed for undergraduate computer science students in their third year. The CS major at Smith requires students to take an intermediate course in each of the following three areas: Theory, Programming, and Systems. This course carries the designations for both Theory and Programming. 
• <your question>

“AI is not Just a Technology” by Claudia Engel, Nicole Coleman

• Thank you very much! What is the intended audience of the course?
• Comment: While I agree with your claim I wonder that in your video you don’t refer to statistics, which in my view cares a bit more about data generating, like e.g. random sampling, measurement than classical ML/AI. Maybe you may want to look at the literature about data literacy education (?).
• <your question>

Today
“Introductory Machine Learning for non STEM students” by Javier Garcia-Algarra

• I think that this is a very good approach! +1 to the competition - this seems to help students to "think with data" (see e.g. https://teachdatascience.com/teaching_programming_tips/#comment-4530154471). I only wonder: why you won't stay with R?
• These students are quite competitive, they love the contest  I say: as long as you are better than me without looking at the data with a simple benchmark model (without preprocession) you will get a good grade in this sub task. This semester >50% were better :-)
• In my experience with engineering students, building your first model from scratch with R or Python is a hard task because of programming details. Agree! There is a Trade-Off. (We stick with https://projectmosaic.github.io/mosaic/articles/web-only/LessVolume-MoreCreativity.html). 
• I have had the same experience with a group of Physics graduate students and they loved the way to build propotypes, despite they understand they will have to translate them right into R or Python

“An Interactive Web Application for Decision Tree Learning” by Miriam Elia, Carola Gajek, Alexander Schiendorfer, Wolfgang Reif

• https://ml.isse.de/dt/
• https://github.com/isse-augsburg/decision-tree-learning-ecml
• Like it! Sometimes I wonder if trees are easier to understand than regression approaches... (For general audience)
• <your question>

“Teaching the Foundations of Machine Learning with Candy” by Daniela Huppenkothen, Gwendolyn Eadie

• https://huppenkothen.org/machine-learning-tutorial/
• I love the idea! +1
• Brilliant! Do you use additonal interactive apps - for visualizing the classifier? But like the handdrawing ;-)
• Great Idea! What is the conclusion of the ethics discussion?
• (Comment: this summer https://teachdatascience.com/ discussed how to include ethics)
• FYI for anyone interested in ethics: There’s a nice article in ACM Trans on Computing Education on Integrating Ethics in ML courses here: https://dl.acm.org/doi/10.1145/3341164 
• For a slightly longer class, I use the AI Blindspot cards to generate discussions: https://aiblindspot.media.mit.edu

“Turning Software Engineers into Machine Learning Engineers” by Alexander Schiendorfer, Carola Gajek, Wolfgang Reif

• https://github.com/isse-augsburg/ecml2020-teach-ml
• <your question>
• <your question>

Breakout groups

Each group will work in a seperate breakout room and discuss one of the topics below. 
You can choose to participate in one of them. 
Please collect your ideas / discussion in the respective etherpad.
The etherpads will be reused -> please don't add information you don't want to share publicly.

1. Education Research https://pad.okfn.de/p/teaching-ml1
   1. What are research projects you know of?
   2. What are research projects that could be done?
   3. Developing a questionaire for students to discover the pedagogical content knowledge (PCK) necessary for teaching concepts in machine learning.
      1. (See also https://twitter.com/HeidiBaya/status/1303675224379056130 )
   4. Which topics are often overlooked when teaching ml? 
2. Opening up educational ressources https://pad.okfn.de/p/teaching-ml2
   1. What open ressources do exist?
   2. What stops the participants in the workshop from making their material available?
   3. What are your experiences with open educational ressources?
3. Learning from others https://pad.okfn.de/p/teaching-ml3
   1. What do experienced teachers (e.g. Carpentries trainers) do that could be used in ML education?
   2. How do you set the goal of a course?
   3. How do you test if you achieved your goal?
   4. How do you balance theoretical foundations and practical skills?
4. Creative teaching ideas https://pad.okfn.de/p/teaching-ml4
   1. Examples of exciting practical projects
   2. What have you always wanted to do as a course/lab?
   3. Which methods have proven helpful when teaching machine learning?
5. <your topic> https://pad.okfn.de/p/teaching-ml5
   1. <your proposed discussion points>

I would like to participate in breakout group... 
(Please sign up with the name that is visible on zoom. Please sign up before the coffee break so we can start right after.)

1.  Education Research
   1. Karsten Lübke
   2. Benedikt Weygandt
   3. Miriam Elia
   4. Heidi Seibold
2. Open Educational Resources and Material Sharing
   1. Oliver Guhr
   2. Cornelia Gamst
   3. Hussain Kazmi
   4. Carola Gajek
   5. Katherine M. Kinnaird 
3. Learning from Others
   1. Peter Steinbach (I might join a tad later as I am overlooking the breakout room assignments from the main room)
   2. Alexander Schiendorfer
   3. Daniela Huppenkothen
   4. <your name>
   5. <your name>
   6. <your name>
4.  Creative teaching ideas
   1.  
   2. <your name>
   3. <your name>
   4. <your name>
5.  <Any topic>
   1. <your name>
   2. <your name>
   3. <your name>
   4. <your name>
   5. <your name>

Guideline for Breakout rooms:

When entering the breakout rooms, we hope that all participants organize themselves. 
However, we provide a rough guideline on how to proceed.

First, introduce yourself with the following ice breaker questions:

• What's your name and why are you here?
• What’s your number one tip for combating distractions when working from home?
• What’s the best piece of advice you’ve ever been given?

Second, discuss the topic of your breakout room. 
Please structure the discussion and work towards a summary that can be presented to the workshop within 6 minutes plus 3 minutes of discussion. 
Designate a speaker that will present a summary of the discussion.

Summary of Breakout rooms
(7 minutes for presenting, 3 minutes for Q&A => https://cuckoo.team/teaching-ml)

1. Education Research
2. Open Educational Resources and Material Sharing
3. Learning from Others

General Feedback
==============
first something that you liked, that you want us to keep doing

• great structure +1+1+1 :)
• very good presenations and discussion, well organized
• I really liked my breakout session group +1
• that this workshop exists +1!
  • this is a crucial point
• clear formulation of objectives and what's expected from attendants
• the preface event +1 :)
• it would be great if you (we?) continue to work on this +1
• the irganization with the Pads and the wonderful inputs!

second something that you didn't like or were irritated by

• test zoom webclient performance before workshop
• for me personally the call for paper was not that clear (i.e. the expected "scientific level" of contribution)
• my breakout group could have been prepared and structured a little better, as it was the main part of the workshop for me