Developing a Carpentries-style Machine Learning workshop

Lightning talk ⚡️

Dr. Jens Brinkmann

The University of Auckland

Tuesday, October 17, 2023

Overview

• To answer the main question “ML-Carp Workshop, what is that?” let’s answer three sub-questions:
  • Q1: What was done?
  • Q2: Who participated and how did they feel about it?
  • Q3: ML-Carp’s future?
• This talk links to upcoming sessions:
  • Upskilling researchers in Machine Learning
  • BoF Session: AI Skill Training Pathway: Bridging Gaps and Fostering Inclusivity

What was done

Background

• Machine Learning (ML) or Artificial Intelligence (AI) are without a doubt hot topics
• Empowering researchers to understand and use ML seems rewarding, but also challenging
• The Carpentries offer a proven teaching-style accounting for participants with limited IT experience
  • Core aspects
    • live-coding
    • positive attitude towards mistakes
    • minimal prerequisites
    • maximising participant involvement
• The CeR (Centre for eResearch) at the University of Auckland acquired funding to further develop and run a series of workshops (also referred to as ML-Carp)
• Two iterations (referred to as Run #1 and Run#2) took place so far. In March and September 2023.

Lesson Overview

Lesson Title	Status	Run #1	Run #2
Programming with Python	Released	Mon, Tue	-
Introduction to Machine Learning with Scikit Learn	Alpha	Wed, Thu	Mon, Tue
Introduction to Deep Learning	Beta	Fri	Wed, Thu

 

• 4h sessions over 4 consecutive afternoons

• Depending on Maturity, sessions were run with small changes or substantially (re)developed.

• One major change was Goolge Colab being used instead of local Python installs.

Lesson Details

Programming with Python	Introduction to Machine Learning with Scikit Learn and Python	Introduction to deep learning
Python Fundamentals	Introduction	Introduction
Analyzing Patient Data	Supervised methods - Regression	Classification by a neural network using Keras
Visualizing Tabular Data	Supervised methods - Classification	Monitor the training process
Storing Multiple Values in Lists	Ensemble methods	Advanced layer types
Repeating Actions with Loops	Unsupervised methods - Clustering	Outlook
Analyzing Data from Multiple Files	Unsupervised methods - Dimensionality reduction
Making Choices	Neural Networks
Creating Functions	Ethics and the Implications of Machine Learning
Errors and Exceptions	Find out more
Defensive Programming
Debugging
Command-Line Programs

Who Participated And How Did They Like It?

• Large sign-up numbers meant mandatory expression of interest (EOI) and filtering
• Analysis follows from two perspectives:
  • Metrics
  • Statements

Participants’ Background

Participants’ OS

Participants’ Feelings - Metrics

Zoom attendance

Participants Promotion/Demotion

Participants’ Feelings - Statements and Adjustments

• observation: At very end of the Runs, questions/participation/interest didn’t cease!

• “Python lesson might be separated”

  • Run #2: Python was removed from the curriculum

• “more breaks”

  • Run #2: better adherence to schedule

• “stick to one coding interface”

  • Run #2: only Colab, VSC etc. was linked to for post-workshop engagement

Note

All statements are positive phrased; overall the results confirm a poasitive appreciation.

ML Carp’s Future

• while specific funding (for lesson preparation and delivery) has ended, the importance of the topic continues to increase
• more sessions are planned
• maintaining momentum using dedicated Slack sessions seems promising
• Feedback about separating the Python lesson is positive

Thank you and Contact details

• Centre for eResearch Website
• HackyHour
• 📨 j.brinkmann@auckland.ac.nz

Developing a Carpentries-style Machine Learning workshop