This is the beating heart of course expectations, policies, and schedules. If something isn't clear, please ask; claiming you didn't know is not an acceptable defense.
Discord
Discord is our primary method of communication. The email you receive from me inviting you to the Discord server will likely be the last email I will proactively send you. While you are welcome to email me, I receive dozens--sometimes hundreds--of emails each day. Discord cuts through the noise, making it much more likely I will respond to questions on the server. Plus, in Discord, you have your student colleagues who can probably help even more quickly.
Grading
There will be five assignments. You are required to turn in three; any more are extra credit.
Everyone is required to sign up for and deliver at least one workshop (students in 6360 are required to deliver two workshops). Any more is extra credit.
Choose one:
- Midterm Exam (Tuesday, Oct 12)
- Final Project (proposal, updates, presentation)
| Assignments | 45% |
| Workshop | 15% |
| Midterm | 40% |
| Final Project | 40% |
There is NO FINAL EXAM for this class.
Assignments
There will be five assignments through the course. Each assignment will be released on Thursday mornings, and will be due by 11:59:59pm two and a half weeks later on a Tuesday.
For every 24 hours an assignment is late, 25% will be deducted from the final grade.
Follow directions precisely in terms of code design; auto-graders will be implemented to run your code! If the auto-grader fails, it may not necessarily be due to a bug in your code, but rather that you did not adhere to the input/output design guidelines of the assignment.
Workshops
Workshops are held most Mondays. Each student is required to organize and lead at least one workshop, most likely requiring workshop presentation teams of 2-3 students each week.
Students taking CSCI 6360 are required to present two workshops.
Workshops are meant to be an opportunity to dig into the nuts and bolts of topics we are covering in lecture by demonstrating working code and how these topics can be implemented in practice. This can take several forms--students are encouraged to be imaginative!
You can demo a new Python package, show a better way of doing something from lecture or homework, or even a live-coding session for how to solve a certain problem (though it should go without saying that you need to practice the live coding ahead of time).
The current schedule can be found here.
Midterm
The midterm exam will be Thursday, October 12. Expect a mix of multiple choice, short answer, and coding questions.
If you opt to do the Final Project instead of the midterm, you do not need to attend class on the day of the midterm exam.
Final Projects
You are encouraged to form teams of 2 or 3. Larger teams, or teams of a single individual, are not forbidden but are discouraged.
There are three components: the proposal, two periodic updates, and the presentation. The proposal is a 1-2 page roadmap of your project, detailing how you plan to do it, any contingencies and outcomes you can surmise, and your teammates. The updates entail two deadlines between the proposal and the presentation where you specify in 1 page what you have accomplished so far, any obstacles you have encountered, how you plan to handle them, and any deviations you anticipate from the plan spelled out in your proposal. The presentation is a 25-minute talk summarizing your problem, motivating your approach, and discussing your results.
Final Exam
There isn't one! Hooray!
Recommended Textbooks
The course has no required textbook. However, there are several recommended textbooks that this course will draw on over the semester. Should you want more information on a topic, these would be good places to start.
| Title | Author(s) | Links |
|---|---|---|
| Elements of Statistical Learning | Trevor Hastie, Robert Tibshirani, and Jerome Friedman | [reviews] [amazon] [pdf] |
| Statistical Learning with Sparsity | Trevor Hastie, Robert Tibshirani, and Martin Wainwright | [amazon] [pdf] |
| Pattern Recognition and Machine Learning | Christopher Bishop | [reviews] [amazon] |
| Probabilistic Machine Learning | Kevin Murphy | [github] |
| Mathematics for Machine Learning | Marc Peter Deisenroth, A. Aldo Faisal, and Cheng Soon Ong | [github] |
| Convex Optimization | Stephen Boyd and Lieven Vandenberghe | [amazon] [pdf] |
| Computer Vision: Algorithms and Applications | Richard Szeliski | [amazon] [pdf] |
| Active Contours | Andrew Blake and Michael Isard | [amazon] [pdf] |
| Probabilistic Graphical Models | Daphne Koller and Nir Friedman | [amazon] |
| Lecture Notes on Spectral Graph Methods | Michael W. Mahoney | [pdf] |
| Statistical Analysis of Network Data | Eric D. Kolaczyk | [amazon] |
| Deep Learning | Ian Goodfellow, Yoshua Bengio, and Aaron Courville | [amazon] [html] |
| Machine Learning | Tom Mitchell | [amazon] [html] |
"A Culture of Honesty"
Here's the short version: don't copy code from the internets or from your friends. Coding is like writing: everyone has their own style, and it's easily recognizable. As such, forgery and plagiarism is also easily sniffed out. Plus, there are tools that do this for us now (this is, after all, a data science class).
If cheating is uncovered, I am obligated to report the incident, no questions asked. As in, I don't warn you beforehand or give you a chance to apologize; the first you'll hear that I think you've been dishonest is from UGA, not me.
The UGA Academic Honesty Policy is the final word on these matters. Lack of knowledge of these policies is not sufficient justification for violations. If in doubt, ask me.
Use of chatbots is not prohibited, but not encouraged either. I won't go into the irony of using an ML-based chatbot specifically not optimized for conceptual correctness to help you ace an ML course, but suffice to say: like any other form of "presenting someone else's work as your own," it's pretty easy to notice. I'll notice. And while I won't mark down for using it (unless it's incorrect, of course), don't expect a glowing recommendation letter from me.
