Signal Processing - July 2016 - 70

Community Teaching Assistants (Rice University)
Conventional residential courses are typically staffed by one
or more teaching assistants, who grade and sometimes create homework and exams, hold office hours, and facilitate
laboratories. The concept is typically also ported over to
online courses such as massive open online courses
(MOOCs). The Rice ELE301x discussion forum quickly
became vibrant in both editions of the course. Moreover, certain students began taking a leadership role when answering
forum questions and offering advice. One of these students,
John Coppens (see https://www.jcoppens.com), a practicing engineer with a wealth of real-world experience, was so
active in the first edition of the class that we elevated him
onto the course team as a community teaching assistant for
the second edition.
This is an intriguing exemplar of how the "openness"
of an MOOC can lead to emergent behaviors that are
unseen in residential courses. Moreover, this is a preliminary indication that the dream of MOOC scalability

Case studies
In addition to the usual analysis and calculation assessment, Rice graduate student Eva Dyer developed a suite
of case studies that challenge students on the core concepts of the class through a series of real-world application
programming exercises. These studies enable students to
explore first hand how signal processing is used in a wide
range of real applications. Case study topics include audio
synthesis using sinusoids, predicting financial time series
using moving average filters, audio synthesis of a clarinet
using attack-decay-sustain-release curves, filter design for
Karplus-Strong string synthesis, and neural spike sorting.
In the second edition of the course, case studies were integrated into the weekly homework.

Office Hours videos
Rice graduate student Raajen Patel prepared videos of worked
examples that give students more insight into the material and
encourage strong problem-solving skills.

Post-MOOC community
Because interest in both editions of the course continued
even after they were closed and archived, we worked to
engage students in an ongoing, post-MOOC community.
Many students were interested in staying in touch with
both course staff and other students. One particularly
useful contribution from 200 members of the community was to work out step-by-step solutions to a number of signal processing problems. These solutions were
then used as test data for a Rice project on mathematical language processing [S2] that aims to automatically
grade mathematical calculations and provide appropriate
70

could in fact be realizable. With so many of the
ELEC301x students having advanced degrees (see the
section "The Class by the Numbers"), we plan to continue to
encourage and reward such positive contributions in
the future.
John Coppen's reflections on ELEC301x provide a number
of insights into why he took such initiative: "I was very interested in the subject, and, over the years, have been studying and implementing small projects for myself. The course
coincided with the first semester of the year, when I normally have a little more time to spare, and the general feel of
the course through its videos and discussion forum was
'open,' 'stimulating,' and 'inviting.' Moreover, reactions to
problems posted to the forum were prompt and helpful.
Minor issues with the tasks and exercises actually forced me
to do more investigation than was called for in the course,
which was a great learning experience. Perhaps MOOC
instructors should leave some such issues on purpose."

feedback to students. For two additional unique elements,
see "Community Teaching Assistants (Rice University)"
and "MOOCs as an Experimental Platform (Rice University)," respectively.

The class by the numbers
We ran an interesting experiment purely by accident. For
the second edition, very close to when standard practice
would dictate that we announce the course (several months
in advance), we were strongly recommended to split the tenweek course into two five-week courses. Accomplishing the
split took several months, and so the course was announced
late (approximately one month in advance). This reduced the
number of registrants significantly (see Table 5). However,
the proportion of students that actually completed at least the
first part of the course was more than twice that in the first
edition. This is evidence for the hypothesis that registration
numbers are not very informative for MOOCs, because many
potential students will register after reading the announcement of a course without ever truly intending to complete the
work required to finish it.

Retention and engagement
The yield of enrolled to passed students was above the average
for edX courses at the times our two editions were offered;
see Table 5.

Demographics
The median age of residential Rice students taking
ELEC301 is 20, but the median age for students taking the
two editions of the course was 27. Interestingly, 30% of
registrants had a high school diploma or less, 39% had a

IEEE Signal Processing Magazine

July 2016

https://www.jcoppens.com

Table of Contents for the Digital Edition of Signal Processing - July 2016