CS2023 Machine Learning CS Core

Todd W. Neller
Gettysburg College

Summary These materials cover the CS2023: ACM/IEEE-CS/AAAI Computer Science Curricula CS Core topics of the Machine Learning (AI-ML) knowledge unit (Computer Science Curricula 2023, p. 71). Using interactive Python Jupyter notebooks, not only are topics introduced experientially with associated exercises, but notebook code examples provide a recipe book for immediate application of principles.

Topics supervised learning (classification, regression), reinforcement learning, unsupervised learning (clustering), fundamental ideas, overfitting, bias-variance tradeoffs, data-preprocessing, missing values, categorical values, normalization/standardization, feature engineering, train/validation/test sets, performance measures, hyperparameter tuning, neural networks, and ethics for machine-learning

Audience Intermediate undergraduate Computer Science students (CS Core)

Difficulty Post CS2 Computer Science undergraduates that are proficient programmers and comfortable with mathematical notation. This does not presume prior introductory AI coverage.

Strengths CS2023's call for universal CS coverage of a great many Machine Learning concepts within 4 CS Core hours is a daunting task. These materials provide aid to instructors seeking to fulfill these curricular goals. It is our hope that such materials will positively advance this important topic area through broad instructor aid if not full adoption.

Weaknesses Code is currently only available in Python. (Ports to other languages are invited!)

Dependencies Students must have intermediate undergraduate programming skill and be comfortable with mathematical notation for definitions of various machine learning problems, metrics, etc.

Variants The offered teaching examples and exercises may be modified and tailored to individual instructor use. The AI-ML CS Core topics are difficult to cover with any depth within 4 hours, yet seeing a minimal treatment here, we hope that instructors are encourage to use and remix this approach so as to personalize and strengthen these offerings.

Project files - See ml-cs-core-*.ipynb files for each of the 4 units.

Instructor Solutions available on request

Summary	These materials cover the CS2023: ACM/IEEE-CS/AAAI Computer Science Curricula CS Core topics of the Machine Learning (AI-ML) knowledge unit (Computer Science Curricula 2023, p. 71). Using interactive Python Jupyter notebooks, not only are topics introduced experientially with associated exercises, but notebook code examples provide a recipe book for immediate application of principles.
Topics	supervised learning (classification, regression), reinforcement learning, unsupervised learning (clustering), fundamental ideas, overfitting, bias-variance tradeoffs, data-preprocessing, missing values, categorical values, normalization/standardization, feature engineering, train/validation/test sets, performance measures, hyperparameter tuning, neural networks, and ethics for machine-learning
Audience	Intermediate undergraduate Computer Science students (CS Core)
Difficulty	Post CS2 Computer Science undergraduates that are proficient programmers and comfortable with mathematical notation. This does not presume prior introductory AI coverage.
Strengths	CS2023's call for universal CS coverage of a great many Machine Learning concepts within 4 CS Core hours is a daunting task. These materials provide aid to instructors seeking to fulfill these curricular goals. It is our hope that such materials will positively advance this important topic area through broad instructor aid if not full adoption.
Weaknesses	Code is currently only available in Python. (Ports to other languages are invited!)
Dependencies	Students must have intermediate undergraduate programming skill and be comfortable with mathematical notation for definitions of various machine learning problems, metrics, etc.
Variants	The offered teaching examples and exercises may be modified and tailored to individual instructor use. The AI-ML CS Core topics are difficult to cover with any depth within 4 hours, yet seeing a minimal treatment here, we hope that instructors are encourage to use and remix this approach so as to personalize and strengthen these offerings.