CSci574 / AI 520 Machine Learning: Classroom Session 08/20/2026
Overview
Derek Harter introduces a graduate machine learning course focused on classic supervised methods, including regression, support vector machines, decision trees, and ensembles, followed by PCA and clustering. He explains the approximate grading structure and walks through the required GitHub, SSH, Docker, VS Code, and Jupyter setup for Python assignments, including how to accept and run the first assignment.
Key takeaways
- The course emphasizes implementation and analysis in Python: timed tests and assignments use notebooks or code, and students apply methods such as regression, classification, and clustering to data.
- The grading plan assigns about 60% to seven or eight programming assignments plus an open-ended dataset project, and 40% to one midterm and one final worth 20% each.
- Using the provided Docker development container avoids many package-version conflicts because it matches the Python and library versions expected by the autograder.
- Students must accept the course organization's GitHub invitation, configure an SSH key, and verify authentication before cloning or submitting assignment repositories.
- Windows Git users should preserve line endings as-is during installation to avoid compatibility problems when working with Unix-based Docker containers.
- The curriculum begins with supervised regression and classification, covers support vector machines and tree ensembles, then turns to PCA and k-means clustering near the end.
Chapters
- Derek Harter waits for students to join, checks the audio, and begins the recorded class after the scheduled start.
- Students should log in to D2L and review course announcements and content.
- The class combines multiple sections, including online and face-to-face students, so the session format may vary.
- Derek Harter identifies Science 355, the high-performance computing lab, as his usual workspace and invites students to arrange meetings by email.
- Students should use the course identifier in email subject lines or send messages through D2L to help route messages correctly.
- Face-to-face sessions are scheduled in Ferguson Hall across from the journalism building, and recordings may be posted for students who cannot attend.
- The course uses Python, NumPy, and scikit-learn; students need programming experience but do not need prior Python experience.
- Derek Harter recommends the textbook's third edition as a longer-term reference, with the course primarily covering its first nine chapters.
- A free Python reference book and early lecture notebooks can help students review loops, functions, and basic syntax.
- Roughly two-thirds to three-quarters of the course covers supervised learning, including linear and logistic regression, classification, and support vector machines.
- Decision trees and ensemble methods such as forests remain useful alternatives to neural networks, depending on the data and task.
- The final portion introduces unsupervised learning, especially principal component analysis for dimensionality reduction and k-means clustering.
- Assignments account for about 60% of the grade, with approximately seven or eight programming assignments and a more open-ended project analyzing a chosen dataset.
- The midterm and final are each worth 20%, for 40% combined; Derek Harter expects to use one midterm rather than two.
- Online timed tests may use IPython notebooks for coding tasks such as loading data, creating visualizations, and running regression analysis.
- Docker containers are recommended because they provide the Python and library versions used by the course autograder, but students may install Python locally instead.
- Students using a local Python installation are responsible for resolving any package-version incompatibilities with assignment requirements.
- Assignments are due Sunday at midnight, generally about every two weeks; students should begin the first assignment early rather than wait until the deadline weekend.
- Students need a GitHub account and must accept the invitation to join the course's 574 organization before working on assignments.
- Derek Harter recommends using a Leo email address for a new GitHub account, though he allows students to use an existing personal account.
- The course uses Classroom 50 as a replacement for GitHub Classroom, and organization invitations may expire if not accepted.
- Students need Git and SSH installed locally; macOS may offer Git through the Xcode command-line tools, while Windows users can install Git for Windows.
- During Windows Git setup, select the option to check out and commit line endings as-is rather than converting them to Windows format.
- Students should configure Git's global username to match their D2L gradebook name and set the Git email to the primary email on their GitHub account.
- Run ssh-keygen and accept the default file location; newer installations will commonly create an ED25519 key pair.
- Copy the public key, the file ending in .pub, into GitHub's SSH key settings, and never use the private key for this step.
- Test SSH authentication from a terminal before cloning repositories; a successful authentication message confirms the key is registered.
- On Windows, Docker workflows may also require starting the OpenSSH Authentication Agent service so the key is available to containers.
- Docker Desktop is recommended for the course's development containers, and students should select the installer for their machine's architecture.
- Mac users need to distinguish Apple silicon from Intel hardware; Windows users may also need to identify ARM versus Intel or AMD hardware.
- After launching Docker Desktop, confirm that the Docker engine is running before trying to create a container.
- VS Code works as the course development environment, but students need the official Microsoft Dev Containers extension to open assignment projects in Docker.
- The extension installs required development tools inside the container, rather than relying on the host system's Python setup.
- Students who already have a personal GitHub account may use it for the course repositories, even if its primary email is not their Leo email.
- Use the assignment invitation link while logged in to GitHub and after joining the 574 organization; accepting it creates a private repository for the student's work.
- Clone the repository with its SSH URL, which requires a working SSH key, and open the cloned directory as the VS Code project root.
- Course repositories are owned by the 574 organization, and students push commits to them so Derek Harter can review and grade their work.
- In VS Code, reopen the cloned assignment folder in a development container; the first build may take time while Docker downloads its components.
- The assignment environment uses Python 3.13 and includes Jupyter Lab; students can open Jupyter Lab in a browser or run notebooks in VS Code.
- If selecting a notebook kernel manually, use the environment under /opt/base rather than a host Python installation.
- Assignment 1 asks students to implement functions such as sigmoid in a Python source file and test them from notebooks; Derek Harter plans to continue the walkthrough at the next Tuesday session.
Transcript excerpt
0:47 Give it a little bit of time here. See if we get some others. Almost ready, I think. What about that?
1:59 So, I have one person on just to give us a moment here. I'm waiting to see if we get some more join us here before I start in earnest. Uh, >> hello, Professor. >> Hi. How's it going? >> Good. How are you? >> Good. So, I guess the audio is okay and everything. >> Yeah. >> Okay. Is there in-person class today or
3:11 >> we've got one one local and and you uh online at the moment. So yeah. Uh just waiting a little bit longer. See if we have some more join us one way or the other. But we'll get started in a minute. I'm actually still setting some stuff up here to >> Yes. >> Sure. like hear the most I think. Yeah.
4:31 I guess that worked. Check some things. Yeah. Hope this blood is like this. Well, um, okay. Yeah, it's five past.
5:47 I'm gonna go ahead and get started here, even though we don't have too many yet. So I am recording this. Um so I'll I'll post these. Uh let's let's jump to the course stuff first. So you know um my plan for today was kind of typical uh first day stuff. We'll go over the syllabus. Uh see if people have questions, things like that. I got a couple of things that um tasks that I want people to be aware of. So, um, so let me So, if you haven't logged in to D2L, make sure you do that. Um, so I've mentioned in some announcements, I'm actually kind of, uh, we've got kind of three different sections combined here. So, the form might be a format might be a little bit uh, strange
This is the start of the transcript. Save this video free for the whole transcript, searchable and time-coded.
In this course
Related notes
Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, Derek Harter.