I was invited to be a volunteer mentor on the Sequence Models course which is a part of the deeplearning.ai Deep Learning Specialization on Coursera. This is a course associated with Stanford University. The course covers Recurrent Neural Networks for Natural Language Processing. I got the invitation by email a few weeks after I completed the 5 courses in the Specialization in May 2018. I did this as a follow-up to the ever-popular Machine Learning course by the same instructor.
I was surprised when I got the invitation as I did not expect this and was not sure what it would entail. This was actually one of the more difficult courses for me since I only briefly looked at NLP prior to taking the course. This was not an issue as they select mentors based on high grades and engagement with the community via the course forums.
In this post, I cover why I decided to accept the offer, mentor training, and my experience so far.
Why did I decide to mentor?
Mentoring can require a significant time and energy commitment. The first few months involve 3 hours per week. Nevertheless, I decided to accept for a few reasons.
It is a growing and learning experience.
I have never mentored an online course before so I decided to keep an open mind and see what the experience would be like. I have done many online courses from sundry providers since this is one of the main ways I stay current and I have been curious about what it is like to create and manage one of these large-scale courses. Hence, I wanted to see what it is like from the perspective of being the mentor versus being the mentee.
I am also relatively new to this particular part of deep learning and I realize that by staying involved it will help me to become a better subject matter expert as well as a better leader.
I wanted to give back to the community.
I have had a lot of help along the way and I know what a huge difference it can make to have guidance when you get stuck. There was a time when I did not have this facility so I do really understand the value of having mentors around. Furthermore, I think you get good karma points from the universe when you help others .
In order to become a mentor, you first have to pass a training course. The invitation comes in the mail and once it is accepted you are enrolled. The course is 3 weeks long but can really be completed in less time. There is also an honors track for people who have been mentoring for a while with extra material to cover. I opted to only take the non-honors track for now.
The course is done in the typical format of a Coursera course. There are lecture videos, readings, discussion prompts, quizzes and peer-reviewed exercises. The content covers information on expectations as a mentor, tests on subject matter knowledge, details on how the platform works like the quizzing system, versioning and sessions, as well as exercises on how to write encouraging posts, give feedback and develop the community by sparking interesting discussions.
Differences from other Courses I have Done
There are some major differences in this course from the ones I normally do on Coursera. I have done quite a few of them so I noticed. Understanding all the different ways things are done as a mentor has taken some getting used to. It is a much more interactive environment than the format I usually follow as a learner.
The Mentor Portal
There is a Mentor Portal. This is a dedicated forum area where you go to get support from the other mentors including the senior mentors. There are mentors who guide the newer ones.
There are several peer-reviewed exercises. These are the first peer-reviewed exercises I have ever done and I had to write a few essays and evaluate submissions from other mentors-in-training. Since I normally do technical courses which have code assignments and quizzes only this was new to me and understanding how this worked had a learning curve in itself.
In my first exercise, I wrote about how I got interested in the subject, and I did 2 pieces on my favorite parts of the course. The first involved a description of Trigger Word Detection. The second a description of how Recurrent Neural Networks fundamentally work.
I really enjoyed this exercise because it required explaining the topics in the clearest way possible and it made me think carefully about how to do this. I actually redid the second description because I felt my first attempt was not so well-done. The grades and feedback appear at a certain date in the future from 3 graders who are also learners on the course. One of the graders really liked my contribution and sent me some very kind words that made my day .
Reviewing other Learners
After submitting my assignments I then had to review 3 assignments from other learners. In this case, it was necessary to pass the course. This was a very strange experience for me initially. The first assignment given to me was not clearly written. It didn’t have paragraphs either so it was difficult to read.
I skipped it since I did not have the heart to give negative feedback on my first try! I grew accustomed to this, however, as I got better at giving constructive feedback and realized that this is how we grow and learn ultimately. There is always the option to redo the assignment.
I enjoyed reading the other submissions because I learned so much from them. The next one I got was about neuroimaging and covered the chemistry involved in the process. The piece used an analogy to a Corolla (the car) to explain the orientations the images are taken from. Another was from a more mature learner who worked on satellites and shared information on the history of Convolutions Neural Nets which was an interesting real-life supplement to the actual course which goes by the same name.
Shot via webcams
One of the first characteristics of this course is that it was not professionally shot like the other courses on the platform. Instead, it was shot via webcam! This has the effect of feeling more personal and friendly. It was like talking to friends on Skype or FaceTime.
They hold virtual meetings via Google hangouts where you can meet other mentors face to face. I haven’t attended one yet but I hope to do so soon.
A mentor can retire at any time by unenrolling from the course. If she wants to restart she can enroll again at any time.
Making it Official
Once you pass the course you receive the completion email welcoming you as a mentor as well as an email for your training course certificate. You get a label next to your name identifying you as a mentor for that particular course and you can introduce yourself and start answering questions on the forums! They send a daily digest which points out posts that need attention. There is also a Mentor Newsletter.
I only recently earned my certificate and am still getting used to this so I will write some updates as I go along about my experience. So far most of the discussions are about errors on the platform (Jupyter Notebooks disconnecting and issues with the grader), problems with assignments and general discussions on deep learning. I certainly have a much stronger knowledge of how the platform and the course works after answering learner questions.
Thank you Coursera for giving me the opportunity to be a part of the mentor community!