Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
Browse files

remove unfinished article that got committed by mistake

  • Loading branch information...
commit 5ca312dd0f1200255521da26531fcceb23b5e4cf 1 parent 20aeb67
@rohit authored
Showing with 0 additions and 36 deletions.
  1. +0 −36 articles/thoughts-on-machine-learning-coursera.haml
View
36 articles/thoughts-on-machine-learning-coursera.haml
@@ -1,36 +0,0 @@
--# title: Thoughts About Machine Learning and Coursera
--# excerpt: After taking the online course on machine learning by professor Andrew Ng I felt like jotting down my opinions about Coursera and online learning.
--# published: 2012-10-30
--# keywords: coursera, machinelearning, learning, online, course, mooc
-
-:markdown
- I remember frantically bookmarking OpenCourseWare course material released by prestigious universities like MIT. So many amazing courses thought by some of the best lecturers complete with lecture notes and homework assignments. There was no longer a limit to what could be learned! Machine learning, compilers, advanced algorithms, quantum mechanics and so much more was accessible. Previously to have access to this matter required either enrolling in an university like MIT or buying expensive books. Unfortunately those resources are still languishing in my bookmarks. It turns out that you need a lot of motivation to do these courses by yourself. This is where Stanford decided to change things. They launched two online courses — machine learning and artificial intelligence. The lecture videos were specially recorded for online viewing taking inspiration from KhanAcademy. There were review questions and programming assignments that had to be submitted within due dates. This seemingly minor changes quite literally changed everything. A lot of people participated in these courses and it was a pretty big success.
-
- After the courses were over two companies were founded — Udacity and Coursera. Both are for profit and aim to offer best courses by some of the best minds. The difference between the two on a shallow level is that Udacity is trying to innovate in the delivery of the content whereas Coursera is trying to emulate the classroom experience. I was still sceptical of both of them because I already had a large list of courses from OCW and if I couldn't do them what are the odds that I would do these? I finally decided to dive in and take part in one of the courses. Machine learning it was. About 10 weeks after the course is over I can proudly say that I completed the course! I am elaborating my views on the course, the new wave of online education and thank professor Andrew because he really deserves it.
-
- ## How Does This Work?
-
- The lecturer will record videos which are usually just a whiteboard being written on with the explanation in the audio. You can imagine recording a lecturer giving a traditional lecture with the camera only focussed on the blackboard where he writes about concepts. The videos are also broken up in small portions of 10-20 minutes each. This allows for consumption at times you are comfortable with. Every week a new set of lecture videos are put up. You watch these videos and they have built in questions to make sure you are paying attention. They usually are very easy and only require that you listen to the lecture. After watching the lecture, depending on the type of course you have review questions based on the weeks lecture and programming exercises. Both of these have due dates. Different courses use different grading systems but usually if you complete them on time you get full credit else a small penalty like 5% is deducted.
-
- It doesn't take you long to realize the quality of the lectures. Having professors of Stanford, MIT, Berkley teach you is quite a special thing. Being small in terms of time you don't need to invest large chunks of your day to watch the lectures. I usually watched the lectures in two batches - in the morning and evening. I dedicated Friday and sometimes Saturday for doing the review questions and the programming exercises.
-
- ## Limitations Of The Medium
-
- I didn't know much about machine learning before taking the course but it doesn't take long to realize that it's mostly mathematics. The original course given by professor Andrew at Stanford is very rigorous involving writing proofs and learning the mathematical concepts involved deeply. However for the online version this had to be trimmed. There are no proofs. In fact professor Andrew actually keeps reassuring you that you don't need to know too much math to get through the course. I have to be honest that it felt weird at first but I'm glad that this was done because otherwise it might have been quite difficult and I wouldn't have completed it. The reason why this was done was because there were thousands of students all over the world with differing backgrounds and I believe they wanted to make the course inclusive allowing different folks to participate and learn. By reducing the emphasis on mathematics it allows almost anybody to learn the basics. It also makes it easy on those organizing the course.
-
- The programming assignments were also limited in that they were not complete. You only had to implement functions within a defined structure. The sample data was already loaded for you, and you only had to implement specific algorithms and sometimes just specific parts of the algorithms. Again I think this was necessary because to allow automatic grading of thousands of exercises limiting the scope was necessary. Many people still struggled with programming exercises as was evident from the discussion forums so I think this was a good balance.
-
-
- ## The Good And The Future
-
- The best part of the course for me was the professor. Listening to Andrew Ng speak is a real joy. You can see that he is really passionate about machine learning and teaching it. I'm really glad that in the process of giving Coursera a try I got a chance to listen to Andrew Ng's lectures.
-
- It also very nice to have high quality lectures recorded specifically for online viewing. Most of the lecture recordings available online are in class recordings and not so nice to watch. While watching the Coursera lectures it felt as though they were made for you and I think that counts for something.
-
- Most importantly of course is that anybody in the world with access to a computer and the Internet can learn from some of the best professors in the world. One not so nice thing was that I couldn't access the course videos on an Android phone. I'm sure it's part of their strategy to have the courses accessible on mobile platforms in the future.
-
- ## Conclusion
-
- I think the biggest take away for me from this course is that Maths is important. Far more important than most computer programmers would like believe. If any of you reading this is still in college please learn as much math as possible and learn it well. The |guy who writes long blog posts| has written a nice piece on what kind of maths to learn. As for me I am going to start improving my maths asap. It's a long road and I intend to be patient and keep trying till I start to get a hang of it.
-
- As for the course and this new wave of online classes, I think they are heading in the right direction. I believe that applying machine learning and statistical mumbo jumbo the Coursera system can give better feedback. A nice video featuring co-founder of Coursera, Daphne Koller about connected learning. It's nice and although I don't necessarily believe in everything said in the video and future of learning looks bright.
Please sign in to comment.
Something went wrong with that request. Please try again.