Why these courses?  The Georgia Tech has a strong nonlinear sciences community. One of the ways we promote this effort is through ChaosBook.org, an advanced, open source online graduate level training platform. With thousands of downloads by students worldwide and user feedback, ChaosBook has continuously evolved and kept pace with some of the latest progress in our research community. Since 2013 we have been enhancing the textbook with video clips, 550 so far. The online, on-demand niche course in an area where Georgia Tech excels are meant to serve graduate student study groups at our sister research groups (for example, groups invited to the Spring 2017 KITP transturb17 Recurrent flows - The clockwork behind turbulence workshop), and to help us recruit the best graduate students and postdocs.


As of January 24, 2022

8901 students had registered

~101,000 course videos views, ~6,000 h watch time

98 partially completed the courses

27 received a “statement of accomplishment”

11 had the perfect 100% score

Course concept: Coursework demands are exactly the same online as for our brick & mortar 2nd year graduate class at Georgia Tech. In the online version of the course, students study our web-based hyperlinked textbook, chapter by chapter. At points in the text where a detailed, in-classroom derivation of a result, or a discussion of the motivation behind a particular concept could be helpful to the students, they can click on a hyperlink, and a short, live lecture video is played.

The videos complement the text, as motivational things one says in a live lecture are often too informal for a textbook. The importance of well-designed problem sets cannot be over-emphasized. For these courses Burak Budanur and Xiong Ding (both physics graduate students at the time) built an online automated grader, and formulated and coded the 16 intensive computational weekly assignments. There are few multiple choices and no peer grading assignments for this course. Instead, student has to code a numerical calculation and enter, as a problem solution, one of the many-digit numbers computed in the process. These thoughtfully constructed problems and Python codes are the main motivator for online discussions, and set the pace for the course.

Institutional context: Administrators do not get it - major new curricular and pedagogical initiatives should be driven, formulated and overseen by those who teach. Existing structures, such as instituional professional education departments can help if a course conforms to their "professional" format: PowerPoint + talking head, recorded in their studio. This technology, however, is not suited to teaching advanced mathematics: the core innovation of our courses is integration of a textbook with live lecture segments. Students responding to an online poll largely agree: 20% preferred GTPE recorded PowerPoint videos, 80% the live lecture segments.


Not only do not admistrators not get it, but they actively interfere with teaching, and they are unaccountable for the damage they cause. By our estimate, the two courses have required 12 full work-months (between me, 2 graduate students, and 5 undergraduate video editors). In our case, three weeks before the start of the 2015 course, two (to us totally unknown) administrators removed the course from Coursera (with 8,000 already enrolled) by a single email, and reneged on the grant that was supposed to pay the students who worked so hard on building the course. So we reinvented online education in 3 weeks flat, and delivered the courses on our own Center for Nonlinear Science open-source platform.


In retrospect, we do prefer the freedom of shaping the course ourselves to the Coursera model. The drawback, however, is that the courses are remain largely invisible to the larger internet audience. Despite our best advertising efforts, we have reached only a small fraction of students the course is intended to serve.