My Education in Machine Learning via Cousera, A Review So Far

The problem with Coursera, IMO, is that despite the promise of free education for all, the very large amount of courses with a broad spectrum of usefulness and difficulty makes it too difficult to navigate and can cost a student hours of wasted time watching videos that may be far beneath or above the student's abilities. I would suggest Coursera creates a pre-exam section for all the courses so the student can have a slightly better understanding of what they are signing up for. Hell, they are promoting themselves as a Machine Learning repository, can't they do a basic suggest-a-course section based on pre-exams?

With that said, I love the idea of Coursera, Udacity, EdX, etc. and what they promote. I am one of those people who, despite being American, had zero chance of ever getting into college due to the quality of my inner-city high school and financial circumstances. Of course, the issue is that if you do grow up in this circumstance, you probably won't be able to take advantage of the free education because you may not have a computer. Regardless, anyone who wishes to partake can join with little more investment than a $200 computer from Walmart, so the financial excuse isn't quite the barrier that it once was.

How much opportunity do any of these courses open up? How much do employers give a hoot about a silly certificate from a pseudo-school that rides on the backs of watered-down Ivy League classes? Employers still take a degree from DeVry more seriously than the scattered layout of MOOC, and until this is changed, there will never be tangible proof or progress in the sphere.

I could tell employers all day that I read, and worked through, SICP and Cormen, but why should they care? They shouldn't because there was no one around to tell me my code is wrong. I can show employers websites I built using exotic languages such as Clojure, but why should they care when I can't pass a basic whiteboard test? I could tell them that I can proficiently write in x y z language, but they don't care and they shouldn't. To all of those people who suggest that employers don't care about your background education and only care about the projects you worked on, I challenge you to point out how many of your coworkers don't have at least something that resembles a Formal Education (TM). Coursera is not a Formal Education (TM). The proof will come if, and only if, one day a student self-learned from Coursera with a large portfolio with self-projects nails a job over a kid from MIT with no portfolio. No one, at this point, can seriously say this is happening and a convincing scale.

For all of the courses I've taken (ML, PGM, NLP, SNA) I've followed roughly the same pattern: follow the course for 50-70% of the material doing all homeworks etc. Take what I've learned and go off do some fun projects. Once I hit a road block return to the course and finish up (usually months after the course is 'officially' done) typically forgoing the homework being I've already created 'homework' for myself.

Compared to my classroom experiences, I've found that this method greatly enhances the benefit of the materials I learn later in the course, as well as allowing me to more strongly reinforce material learned earlier. This is also something that physical classes simply cannot emulate. I do find it a little funny that I've never officially completed a coursera course, despite having covered much of the material very thoroughly. Hopefully as these courses continue we'll start to see them adopt patterns that extend beyond replicating the limitations of a physical classroom online.

  ML is the general introductory course.
  PGM is more about applied knowledge.
  NN is a specialized and very good one.

There is also Berkeley CS188.1 on edX, which covers search.

Of course, just courses are not enough. You need textbooks, AIMA, PAIP and PGM are highly recommended.

But most importantly you need practice. One cannot learn how to swim, ride a bicycle or wrestling just by reading books and watching movies.

Very few regions of your brain learn something just by watching and listening. Those which are active when you actually do something must be trained by doing. It is how you will form your own intuitions, timing, etc. Non-verbal parts of the brain are doing the most of the job. Neural networks need to be trained on real data to learn its behavior.)

This simple idea is a cornerstone of MIT approach in teaching CS. They do lots of projects and labs, working in groups. This is real training. Recent version of 6.01 is a good example.

So, start your own project or join some Open Source. Then you will get real practice, training and feedback from other people. It is not meeting that matters, it is practice.