I want to learn Machine Learning. Can anyone please suggest me some MOOCs or any other relevant websites where I can learn it?
Udacity has a course about machine learning. They even offer statistics courses for the ones, that aren’t that proficient in mathematics or - at least - not in that field.
I know there’s one on Coursera but I assume that course will have a higher difficulty-level.
I did not take the Udacity course about Machine Learning, so I can’t really say something about the difficulty but I enjoyed their Git & GitHub course for example. Just try it out and maybe it’s the right thing for you
Have a nice day!
Machine learning can be very very mathematical. The key question is: how much math do you want/can you handle?
How are your math skills re: linear algebra, partial differentiation, basic statistics? Some courses say they don’t need much in the way of math skills, but you need to check, especially if you are paying (caveat emptor)!.I would say, for ANY course, basic linear algebra and basic statistics are an absolute must. You don’t need to know the algorithm for inverting a matrix, but you must at least know what that means (especially for non-square matrices).
If your math skills are not decent, you will merely be running programs you don’t understand. Also, are you proficient in using matrix manipulation systems, as MatLib, Octave, Numpy.SciKit (python)? You might be able to learn during the course, but you can’t do machine learning without this kind of program.
The Stanford/Coursera course is excellent but quite mathematical.
If you can at least understand what this guy is talking about
you’ll be OK. He tries to skip the hard math. See how you like it.
How did you learned it? Can you please tell me your sources?
Thank you for replying
I know maths pretty much but the thing that I am worried about is that I don’t know Scikit, numpy or any of
the programming language that you have mentioned.
Can you please tell me about the language that the Stanford/Coursera course uses?
The course uses MatLab or Octave. Octave is essentially a free version of MatLab. If you’re good in math, either one will be straightforward to learn. The course may be closed to new enrollment now, or soon at least.
Andrew Ng’s course on coursera in Machine Learning is far and away one of the best educations you’ll get on the matter.
it helps to already undestand some linear & matrix algebra but it’s not absolutely required.
Andrew Ng is a great teacher and makes it easy to understand from the beginning. By the end of the coruse I successfully completed the handwriting recognition task.
the code is mostly written for you in Octave/Matlab and you just have to code the vital pieces.
Isn’t Matlab is outdated? I mean everybody only talks about python these days.
If you don’t have a background in math, I’d recommend getting familiar with:
- Basic statistics: your regular mean, mode, averages, as well as regression etc.
- Linear algebra: Matrix multiplications, vector algebra
- Calculus: At the very least single variable differentiation, but I’d do a basic course
Once you’re done do Andrew Ng’s excellent Coursera course.
I did the last semester, and solved everything in Python, which was an amazing learning experience. However, it’s going to need a ton of work if you go that route, and you can’t officially finish the course, since you’re supposed to code in Octave/Matlab. It’s completely fine to learn some Octave to do this though. You’ll pick up the same concepts in Python easily later.
Also someone made excellent notes of this course, and I used them a lot.
Then I’d go on to Kaggle. It’s an excellent resource for data science competitions and has tons of really smart people writing solutions and tutorials for cutting edge ML stuff. I’d also try making a project that utilizes the power of ML.
Of course, this is barely scratching the surface, but it’s how far I’ve gotten in a few weeks of doing ML. Tons of fun if you are relatively familiar with the math (which should be simple-ish if you’ve done high school level math and take a refresher).
i had no experiene with matlab and only very minimal exerpience wiht python when I took the course the first time. (I’m going through it a second time).
I don’t think the language should deter you in this case, and professor Ng actually discusses the choice of language in one of the introductory courses, saying that he’s taught it in many languages, and students picked up on the ML concepts fastest when he taught the coruse in matlab/octave.
There is really no better instructor for machine learning. Stretch yourself, learn a lot about ML, and transitioning to numpy/python should be no problem at all.