UBC CPSC 330: Applied Machine Learning (2023s)

This is the course homepage for CPSC 330: Applied Machine Learning at the University of British Columbia. You are looking at the current version (May-June 2023). Some of the previous offerings are as follows:

• 2020w1 by Mike Gelbart
• 2022s    by Mehrdad Oveisi
• 2022w1 by Varada Kolhatkar
• 2022w2 by Giulia Toti, Mathias Lecuyer, Amir Abdi

Instructor

Mehrdad Oveisi

• [email protected]
• LinkedIn.com/in/oveisi
• Google Scholar
• Office hours:
  • When: One hour after each class. For clarity, that will be potentially four hours per week. I will leave once there are no more questions.
  • Where: Office hours are held in the same classroom after each class.
  • Who: Students form both sections are welcome to attend all office hours.

Class Schedule

Section	Day	Lecture	Office Hour	Location
911	Tue, Thu	13:30 - 17:00	17:00 - 18:00	FSC 1005
912	Wed, Fri	13:30 - 17:00	17:00 - 18:00	DMP 110

Course Coordinator

Jeffrey Ho

• [email protected]

Please email Jeffrey Ho at the above email address for all administrative concerns such as CFA accommodations, extensions or exemptions due to sickness or extenuating circumstances.

License

© 2021 Varada Kolhatkar, Mike Gelbart, and Mehrdad Oveisi

Software licensed under the MIT License, non-software content licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) License. See the license file for more information.

Important links

• Syllabus / administrative info
• Calendar
• Course GitHub page
• Course Jupyter book
• Course videos YouTube channel
• Canvas
• Gradescope
• Piazza (this is where all announcements will be made)
• Setting up coding environment
• Other course documents
• iClicker Cloud (available in Canvas)

Compact course schedule (tentative)

IMPORTANT NOTE
As a general rule, summer terms are quite compact and thus time management is crucial to keep up with the course content and the deadlines. More precisely, based on the university calendar, the number of Teaching Days is 63 in winter terms and it is 28 in summer terms. That means there will be 2.25 (63÷28) times more content to learn per week, and 2.25 times faster pace for the homework due dates. In other words, you are expected to learn and deliver the same amount of work compared to winter terms, but do it 2.25 times faster! For this reason, time management is of utmost importance in order to succeed in the course.

The following chart is a very compact version of the course tentative schedule.

 

The following sections provide for more detailed course schedule.

Deliverable due dates (tentative)

Assessment	Due date	Where to find?	Where to submit?
Syllabus quiz	May 19, 13:00	Canvas	Canvas
hw1	May 20, 13:00	Github repo	Gradescope
hw2	May 23, 13:00	Github repo	Gradescope
hw3	May 27, 13:00	Github repo	Gradescope
hw4	May 31, 13:00	Github repo	Gradescope
Midterm	June 05, 13:30 to 15:00	Section 911: LIFE 2201 Section 912: BIOL 1000	Canvas
hw5	June 08, 13:00	Github repo	Gradescope
hw6	June 13, 13:00	Github repo	Gradescope
hw7	June 17, 13:00	Github repo	Gradescope
hw8	June 22, 13:00	Github repo	Gradescope
Final exam	Jun 27, 08:30	Final exam schedule	Canvas

Lecture schedule (tentative)

Lectures:

• The lectures will be in-person (see Class Schedule above for more details).
• All lecture files are subject to change without notice up until they are covered in class.
• You are expected to watch the "Pre-watch" videos before each lecture.
• You are expected to attend the lectures.
• You will find the lecture notes under the lectures in this repository. Lectures will be posted as they become available.

#	Date	Topic	Assigned videos	vs. CPSC 340
1	May 16,17	Course intro	📹 Pre-watch: None Recap video (after lecture): 1.0	n/a
		Part I: ML fundamentals and preprocessing
2	May 16,17	Decision trees	📹 Pre-watch: 2.1, 2.2 After lecture: 2.3, 2.4	less depth
3	May 16,17	ML fundamentals	📹 Pre-watch: 3.1, 3.2 After lecture: 3.3, 3.4	similar
4	May 18,19	$k$-NNs and SVM with RBF kernel	📹 Pre-watch: 4.1, 4.2 After lecture: 4.3, 4.4	less depth
5	May 18,19	Preprocessing, sklearn pipelines	📹 Pre-watch: 5.1, 5.2 After lecture: 5.3, 5.4	more depth
6	May 23,24	More preprocessing, sklearn ColumnTransformer, text features	📹 Pre-watch: 6.1, 6.2	more depth
7	May 23,24	Linear models	📹 Pre-watch: 7.1, 7.2, 7.3	less depth
8	May 25,26	Hyperparameter optimization, overfitting the validation set	📹 Pre-watch: 8.1,8.2	different
9	May 25,26	Evaluation metrics for classification	📹 Pre-watch: 9.2,9.3,9.4	more depth
10	May 30,31	Regression metrics	📹 Pre-watch: 10.1	more depth on metrics less depth on regression
11	May 30,31	Ensembles	📹 Pre-watch: 11.1,11.2	similar
12	Jun 1,2	Feature importances, model interpretation	📹 Pre-watch: 12.1,12.2	feature importances is new, feature engineering is new
13	Jun 1,2	Feature engineering and feature selection	None	less depth
	Jun 5	Midterm
		Part II: Unsupervised learning, transfer learning, different learning settings
14	Jun 6,7	Clustering	📹 Pre-watch: 14.1,14.2,14.3	less depth
15	Jun 6,7	More clustering	Post-lecture: 15.1, 15.2, 15.3	less depth
16	Jun 8,9	Simple recommender systems	None	less depth
17	Jun 8,9	Text data, embeddings, topic modeling	📹 Pre-watch: 16.1,16.2	new
18	Jun 13,14	Neural networks and computer vision		less depth
19	Jun 13,14	Time series data	(Optional) Humour: The Problem with Time & Timezones	new
20	Jun 15,16	Survival analysis	📹 (Optional but highly recommended) Calling Bullshit 4.1: Right Censoring	new
		Part III: Communication, ethics, deployment
21	Jun 15,16	Ethics	📹 (Optional but highly recommended) Calling BS videos Chapter 5 (6 short videos, 50 min total) The ethics of data science	new
22	Jun 20,21	Communication	📹 (Optional but highly recommended) Calling BS videos Chapter 6 (6 short videos, 47 min total) Can you read graphs? Because I can't. by Sabrina (7 min)	new
23	Jun 20,21	Model deployment and Conclusions		new
24		(optional reading) Stochastic Gradient Descent
25		(optional reading) Combining Multiple Tables