Any self-respecting social media savvy professional knows that LinkedIn endorsements are the result of a hideous gamification experiment gone wrong (on many levels), except when they think it is a straightforward abuse of human psychology. Some computer programmers even try to give it a backlash by writing automated scripts to endorse profiles with totally absurd ‘skill sets‘.
On the other hand, in some unexpected cases, these endorsements can be very motivating, which is what happened to me a few months ago, back in October, 2013. To cut a long story short, when I came across the following endorsements by some of my LinkedIn contacts, my reaction was something that even surprised me:
It went like that: “Machine Learning! Should I accept that endorsement? I mean, I did small projects related to machine learning, such as Poor Man’s TV program Recommender that utilized Support Vector Machines, and a cross-cultural and cross-domain recommendation system using a semantic graph database such as AllegroGraph; but apart from an AI course that I had while studying for my cognitive science degree, I haven’t taken any Machine Learning course. On the other hand, Andrew Ng’s famous Machine Learning course at Coursera is about to start, so maybe that’s a nice opportunity! Why not? If I can finish the course successfully, then accepting such an endorsement will be a bit meaningful, at least from a practical, or academic point of view.”
With such thoughts firmly ingrained in my mind, I started the course, which I had not been motivated enough to complete when it was first offered. I was curious whether I would be motivated enough this time. The result turned out to be positive, even more than I expected: After 10 weeks full of GNU Octave coding (inside GNU Emacs) for linear regression, logistic regression, neural networks, support vector machines, clustering, dimensionality reduction, anomaly detection, recommender systems, and digesting tons of practical advice, as well as examples for machine learning applications, I finally finished the course and more importantly had a lot of fun during the learning process. Even though the course was not very heavy on theory, that did not mean it was easy: the biggest challenge was to keep myself motivated and find enough time; after all, time management theories sound less plausible, even less than string theory, when you are working full-time, married, and have kids. To make things a little more difficult, I was also taking another online course, Tales from the Genome, an introductory genetics course (as well as a physical-class based, traditional French course, oh mon Dieu!) and I wanted to complete them, too. All of this would be simply impossible without the utmost support from my wife and son, who put up with me during many late night and weekend coding & review quiz answering sessions.
A few days ago Coursera sent an e-mail announcing that the results and certificates were ready; checking the web page made me realize that now I can go and accept that ‘Machine Learning’ endorsement:
The slide show below documents some of the fun I had during this class. Many thanks not only to Andrew Ng, but also to people who run Coursera, as well as developers who gave us Octave, Emacs and GNU/Linux. (By the way, Octave finally has an official GUI, not that I care much, enjoying Octave-Emacs integration, but nevertheless, that’s good news.)