By Harry Keller
Editor, Science Education
[Note: Harry, who holds a BS in chemistry from the California Institute of Techology and a PhD in analytical chemistry from Columbia University, is sharing his first MOOC experience in this series. See part 1, 2, 3 and 4. -Editor]
The End of My First MOOC Journey
As an audit student, my motivation to finish is not great. I have to ask if drop-out rates mean not taking the final exam because I never intended to punish myself that way when I began. I already have a very good doctorate in science and have no incentive to acquire more initials after my name today.
The last two lectures of the course were both quite technical and, once you penetrate the jargon, quite illuminating. They brought me to the state of the art in machine learning: SVM and RBF.
I was introduced to machine learning decades ago and was interested enough to obtain a book on it, the only book available at the time. This book was heavy on theory and not very useful to a practitioner. Machine learning in those days consisted of just one approach, known today as PLA or Perceptron Learning Algorithm. It was adequate for simpler learning situations and was adjusted to work in others but was not really sufficient.
The machine learning universe expanded with the advent of neural networks, a learning concept that sprang from biological work at the time. Neural networks are still used but not very often.
Today, support vector machines (SVM) and radial basis functions (RBF) represent where machine learning lies. The former requires an advanced mathematics known as quadratic regression (QR). The latter only requires matrix manipulation and iterative processing, both readily available with software.
With this information, I’m ready to apply machine learning to problems when I find them. I expect that my business will eventually lead to analyzing data with unknown algorithms — in other words, to machine learning.
I found this MOOC to be better than a book and less good than a real course, even a lecture/reading/quiz course. The advantage is that I could listen to the lectures at all sorts of odd moments and even break them up into small pieces to listen to when I could, ten minutes here and half an hour there.
I do not have time to complete the last homework assignment before the deadline. In my business, we are still in the process of completing details, such as performance enhancement, of our new roll-out of our HTML5 version. We are also in the midst of several deals worth hundreds of thousands of dollars and that cannot be ignored or postponed. My personal and possible professional interest in the course cannot compete with crucial business time.
I may be a poor example as a MOOC test case or may be more typical than I assume. I have no way of comparing my situation with the average MOOC student. Throughout the course, I continually had the nagging sense that it could be better. The lectures were well done, although not entertaining. The homework forced you to think deeply and was not at all obvious. As a traditional course, it was excellent and also quite tough.
I hope that the MOOC concept develops into much more that this simple moving from here to there of traditional courses. That movement reduces the effectiveness of the course but increases its reach. The reduced effectiveness restricts the reach from what is possible, however. The discussion groups did not solve the problems of being online instead of face-to-face.
I am experimenting with some ideas to bring real value to the online part my online hands-on science labs now. These ideas use capabilities uniquely available in this format. It’s exactly this sort of thing that will make online education soar. If MOOCs do not incorporate such ideas, they’ll just become extinct as other initiatives outperform them in the Darwinian space of the business of education. Adapt or die.
Filed under: Uncategorized |