Course Review: The Data Scientist's Toolbox

Course Name: The Data Scientist's Toolbox
Platform: Coursera
Rating: 2/5

What the data scientist's toolbox is aimed at, I cannot tell you. Well, the course description says it gives you an "overview" of things, but plainly put, to me it seemed like a course put together so they could offer the 9 courses neatly in bunches of 3 (3x3). The other two courses flanking it are much, much more extensive and demanding.

This course is the equivalent of a walk in the park - no, wait, it's more like being pushed in a pram in the park. The project is a joke, especially if you know Git. All you have to do is create a repo and some other basic tasks, like install RStudio. The quizzes didn't require you to work with any data sets (unlike the other two courses that were offered simultaneously). The videos... well, after around 16 videos in the first week, we get 8 in the second and a paltry 4 in the third. An entire week was devoted to Github and Markdown, both covered at a little more than surface depth. To add to the lack of concrete content, we don't even get a 4th week (unlike the other courses in the "Data Specialization" offering). Clearly, content in this course is lacking.

For all that they did wrong, however, learning about confounding variables was interesting. I'll still have to go through the forums again, and evaluate peer projects, but I doubt I'll find any positives there. At best I can say the video lectures were well produced. I'm trying, I really am...

I'm shocked that they have the audacity to charge ~50$ for this course (though I did it for free). The only plus point here (looking at it in a selfish manner) is that you can easily complete this course and walk. Not much of value is mentioned here, and I would say for data science, you'd be better off trying out other courses instead.

Head First Object Oriented Programming and Design

Urgh. What a book! I'm finally done with this, and only because I decided I'm done and not because I'd read every page of the book. Head First's books are usually a pleasure to read (or at least they were) because things are written in such simple terms that you know you can understand anything. Either their books have changed or I have, because this particular book was just too (too) simple to retain your attention.

That design is a subject that draws me to computer science makes me want to learn how to get better at it. It's always a pleasure to see Norvig's clinical, clearly and carefully crafted programs, and to learn the art that is clearly inherent in the task of writing programs would be great. Sadly, I've taken a strong dislike to huge, chunky classes being thrust into your face as introductions to the problem of design. Head First's book had a lot to say on what I would call pseudo science (or soft science) like requirements and use cases, but I think it did little justice to exactly how to design well.

Reading the book, I couldn't help but think that most of what I was reading was just fluff created to justify the use of buzzwords like "Domain analysis", "Requirement specification", etc. These things seem like very obvious things to do (a to-do list anyone?) and makes me question the value of design patterns and excessive OOPing as I'd call it.

With this book, I'm done for now with this field (I had planned on reading Design Patterns earlier) and instead will move on to something more "concrete".

Learning update

While I'm bogged down by the swampy marshland that Combinatorial Game Theory inhabits, I have tried making progress in other fields.

I've struggled through Head First's Object Oriented Analysis and Design - not so much because of an inability to understand the content (Head First books are rarely too tough to comprehend), but because of the verbosity of its content. The book features redundancy in abundance, and at times feels like it is for kids ages 8-10. Moreover, my aversion to the topic of Design Patterns made it even tougher to plow through the book. To my great satisfaction, a few comments (answers) on Quora stated that the whole idea of "Design Patterns" had been reduced to a fad and its importance, overplayed. This was supported by providing a few snippets from the book Coders at Work.

I've also been working on the field of data science, analysis and machine learning. Progress in this is easier, as learning is easier to move ahead when you are learning compared to when you are reading (not true if the book is interesting). For this, I've been following a few online courses on Coursera, EdX and Udacity.

I also learned about PAC files today, which unfortunately, I cannot get to work with Firefox (a file that works in Chrome...).

I think I need to blog a bit more regularly about any work I do, if I am to use this as a tool for discipline and motivation.

The inevitable gap

And my discipline gives way. I knew it wouldn't take long, but hoped it would and now I task myself with reigning it in.

Recently, I've been trying to delve into the field of combinatorial game theory. Conway's On Numbers and Games was too "deep" for me, so I've tracked back and reviewed Morley's wonderful succinct lectures on the topic. Currently I'm ploughing through a course on MIT. Along with this, I hope to complete a few problems on topcoder and similar sites relating to the topic, after which I should be able to turn to Conway once again (maybe using this as a walking stick along the way).

It looks like I may have to read up about groups, fields and rings too. I hear that is called Abstract Algebra.

Conway's approach is quite fascinating to discover. Developing a number system from games? Ingenious, no less (or should I say surreal?). The whole idea of games that aren't numbers along with their generation is a little tough to wrap your head around though. Have to get to Knuth's "fictional" book on surreal numbers as well.