Digital world evolved - machine learning powered

[2018-05-31 Thu 11:57:10] event: PyCon APAC Edu summit speaker: Norman Lee, SUTD

introduce machine learning in first year python course

The course is “The Digital World”

• raspberry pi
• thymio robot
• google firebase
• propose their own 1D and 2D project

Stucture of The Digital World

• programming concepts (six weeks)
• OOP (two weeks)
• intro to machine learning (one week)
• GUI (Kivy)

Flipped classroom

• students work on mini project during the class
• in groups

Lessons from introducing machine learning

• inconsistent terminology in textbooks
  • follow the terms in scikit-learn
• chose the breast cancer dataset
  • 30 numerical features with binary categorical target
• prior to class
  • videos
  • pre-class activity
  • online quiz
• in-class problem set
• what we want the students to learn
  • the process
  • understanding the dataset
  • model for numerical prediction
    • linear regression
  • model for categorical prediction
    • k-nearest neighbours
    • easier to describe and explain
  • with python, students learn
    • jupyter notebook
    • matplotlib pyplot
    • scikit-learn
    • numpy

Lessons learnt

• code is short, concepts are deep
• numpy arrays are different from regular lists
• steps in machine learning
  • why we need to partition the data into three sets
• jupyter notebook
  • different from what they used before to program
• find a dataset for numerical prediction
• introduce tools earlier