I didn't like Ng's course on ML too much, it is very good if you just want to implement the algorithms, but it gives no insight on why they work.

I enjoyed far more Tom Mitchell's lectures, they gave me a much better understanding of the algorithms described:

http://www.cs.cmu.edu/~tom/10701_sp11/lectures.shtml

There are some comments on HN about it

http://news.ycombinator.com/item?id=3199718

Really? I found descriptions of Linear and Logistical Regression, as well as the descriptions of Support Vector Machines to both explain very well why they work. The Neural Networks description could have been a bit more thorough, but I still understand them much better now than I did before.

This is a great tip. Actually, I was kind of hoping that Ng's course would be focused more on implementation and less on the why, not that the theory is not important, it's just that I have found a lot of other great places on the web to get the theory for free (like the lectures you mentioned).

great! Thanks.

I am actually already following ML from Harvard. http://www.seas.harvard.edu/courses/cs181/extension/

Seems these videos are also great.

I think the rationale is that - at least as Prof Ng articulated in another ML course he taught this quarter - videos following a lecturer as he teaches on stage are actually _less effective_ than videos which just follow slides w/ audio narration. Another researched case was video lectures with the teacher's face in the corner vs. video lectures with only narration. In which they found that student's information retention was even better without _any_ on-screen presence at all. (Surprising, I know . . . I wish I had a citation, but I'm really just coming in on defense of the department's pedagogical choices here.)

Here are the classroom lectures for ML.

http://classx.stanford.edu/ClassX/system/users/web/pg/view_s...

Also, the classroom lectures for the upcoming algorithms course are already up.

http://openclassroom.stanford.edu/MainFolder/CoursePage.php?...

Thanks a lot for the link. Good way to review and compare with online class.

Thank you !Is there the original AI class too?

I partially agree. Complete live sessions are easier to watch when the professor is good and prepared. At the same time, Stanford's format had definite advantages in that it's natively interactive and indexed. If I want to review a concept, I can simply click a link to the appropriate part of the lecture. They also have better sound and image quality.

The ML class is good when watched at 1.5x.

Personally I find that Professor Ng manages to instill a lot of energy into the ml-class.org lecture videos.