This introductory computer science course in machine learning will cover basic theory, algorithms, and applications. Machine learning is a key technology in Big Data, and in many financial, medical, commercial, and scientific applications. It enables computational systems to automatically learn how to perform a desired task based on information extracted from the data. Machine learning has become one of the hottest fields of study today and the demand for jobs is only expected to increase. Gaining skills in this field will get you one step closer to becoming a data scientist or quantitative analyst.

This course balances theory and practice, and covers the mathematical as well as the heuristic aspects. The lectures follow each other in a story-like fashion:

What is learning?

Can a machine learn?

How to do it?

How to do it well?

Take-home lessons.

Syllabus

The topics in the story line are covered by 18 lectures of about 60 minutes each plus Q&A.

by
Gregorycompleted this course and found the course difficulty to be hard.

CS1156x: Learning from Data is a 10-week introductory machine learning course offered by Caltech on the edX platform focused on giving students a solid foundation in machine learning theory. Major course topics include the feasibility of learning, linear models, generalization, VC dimension, overfitting, regularization and validation. The course also covers several common machine learning algorithms including the perceptron, linear regression, logistic regression, neural networks, support vector machines and radial basis functions. As a theory-heavy course, much time is devoted to mathematical…

CS1156x: Learning from Data is a 10-week introductory machine learning course offered by Caltech on the edX platform focused on giving students a solid foundation in machine learning theory. Major course topics include the feasibility of learning, linear models, generalization, VC dimension, overfitting, regularization and validation. The course also covers several common machine learning algorithms including the perceptron, linear regression, logistic regression, neural networks, support vector machines and radial basis functions. As a theory-heavy course, much time is devoted to mathematical reasoning and the math behind various machine learning concepts and algorithms. You need a strong mathematical background, including knowledge of linear algebra and calculus, to understand everything in this course. You also need the ability to program in some language that allows you to perform matrix and vector operations. The course provides a temporary MATLAB license and forum support for MATLAB; many students also used R and Python.

Learning from Data is different from most MOOCs in that it isn't optimized for the web. Course content consists of 18 full-length lecture videos recorded on the Caltech campus, each spanning about 75 minutes including 10-15 minute Q&A sessions. Two lectures are posted each week for 9 weeks along with PDFs of lecture slides and 8 homeworks that each consist of 10 multiple choice questions. There are no in-video quizzes or interactive exercises, as the course is basically an online port of the on-campus course. It requires a high level of motivation and attentiveness get through two very dense 75-minute lectures each week and despite being multiple choice, the homework problems can be very time consuming since many require programming. You get 2 attempts at each question, but each attempt is worth half of your grade, so guessing based on your intuition can be costly. The final is an untimed test that is just like the homework except that it has 20 questions. You need a total score of 50% to earn a certificate.

Although Learning from Data isn't in the typical MOOC format, the professor is a skilled lecturer and manages to keep the lengthy lecture videos engaging. The lecture slides are packed with useful information and the forums were very helpful; students were active in helping one another and the professor was very active on the forums even though this wasn't the first run of the course. The homework questions reinforce the material more than you would expect from a 10 question multiple choice quiz if you take the time to understand the question and answer carefully.

Overall, Learning from Data is a great course that emphasizes theory, but often has practical implications. The level of mathematical maturity it requires will be barrier for some students, although you can still get something of this course if you don't understand all of the math. If I were taking this course as a student on campus I would probably rate it 5/5, but I think they missed some opportunities to make it truly excellent MOOC by failing to adapt it for the online the audience. This course will give you a deeper understanding of machine learning than other intro MOOCs on the same subject, but if you're more interested in learning practical tools and applying machine learning consider taking MIT's Analytics Edge on edX or Coursera's Machine Learning course.

by
Ronnycompleted this course, spending 12 hours a week on it and found the course difficulty to be hard.

Professor Yaser Abu-Mostafa created an exceptional course and provides it for free to everyone who wants to take the time and effort to dive into this excellent material. His domain and instructional skills are top of of the bill and the world should be thankful he makes this available to millions, skills which belong to the most wanted in industry today.

I am a bit of a MOOC addict having finished more than 50 MOOCs so far. This machine learning course belongs to the top3, if not the top1 course I followed. I very much appreciate it that the MOOC mirrored the in-class semester co…

Professor Yaser Abu-Mostafa created an exceptional course and provides it for free to everyone who wants to take the time and effort to dive into this excellent material. His domain and instructional skills are top of of the bill and the world should be thankful he makes this available to millions, skills which belong to the most wanted in industry today.

I am a bit of a MOOC addict having finished more than 50 MOOCs so far. This machine learning course belongs to the top3, if not the top1 course I followed. I very much appreciate it that the MOOC mirrored the in-class semester course and not watered it down to something simpler to attract more people, a regretful technique applied too much on todays MOOC platforms. The Prof's contributions on the discussion forum together with the TAs are exceptional, they help you to push through the difficult moments.

I knew upfront that this would be a tough one, 10+ hours per week combining with a heavy loaded fulltime job and family life is not obvious, but I am so glad I stayed disciplined and did it.

So be prepared, you'll need to be reasonable confortable with calculus, linear algebra, basic stats and a data science friendly programming language (python, R, matlab, ...) to achieve a good result on this course. Free up 10-20 hours weekly to focus on the course material and the exercises. If you think this is too demanding for a first encounter with machine learning, take a look at other MOOCs first, for example Andrew Ng's Machine Learning course on Coursera or The Analytics Edge course on edX, both are somewhat less demanding but evenly engaging. But be sure to come back to this course afterwards!

If a nobel prize for education would exist, Yaser Abu-Mostafa would be my number one candidate!

The best Machine Learning class available for free online, period (I also took Coursera/Stanford's). This class will make you understand very well the principles underlying machine learning. You will do some programming assignments as well but the goal of those assignments is for you to understand what you are doing and why you do it (vs implementing some textbook algorithm for the sake of it). It has the right balance theory/practice. Serious students will love it.

by
Harishcompleted this course, spending 14 hours a week on it and found the course difficulty to be hard.

I would highly recommend this MOOC to anyone who is interested in machine learning. Every week consists of two lectures (each an hour long) and a problem set of 10 questions. The duration of lectures seem to be long in number only but once you start listening to it it just flies because the professor has super style of breaking complex this into simple ones . Prof Yaser beautifully explains each and every concept in depth that even some who is new to this field like me, will enjoy the course. Also this is one of the very few courses which has a very active discussion forum with great TAs, community TAs and fellow students. More than all, the Professor himself participates very actively in responding to questions which is very kind of him.

Overall I would say that the learning experience was extraordinary and definitely worth the time.

The lectures are engaging, and the homework and exams very challenging. Several topics presented have made me excited to pursue them further.

The only downside was that homework and exam problems were multiple choice, one try only. When I got one wrong, I knew there was something I didn't understand, so I went back and learned from my mistakes until I got the right answer. Still my scores were lousy.

If you take this and care about your scores, be sure to follow the discussions in the forum (if only I had!).

This MOOC is very technical: it requires knowledge of matrix algebra, calculus, and programming skills. The lectures are great, although some weeks are too theoretical to my taste, and an application is not clear. The professor ,no doubt, is an expert and delivers material in a good pace.s But you have to stay on schedule, missing one week is enough to put you in trouble of making up. Homework takes a while. Plan to spend about 12 hours a week working on it.

by
Bobbycompleted this course, spending 11 hours a week on it and found the course difficulty to be very hard.

Groundbreaking course derived from the original CalTech telecourse, that introduces you to the theoretical and mathematical concepts behind many machine learning algorithms and models. Provides with you no background material, so you must be competent in your prerequisites before beginning. In addition to having a concrete understanding of Statistics, Probability, Linear Algebra & Calculus be prepared to effectively use at least one object oriented or functional programming language.

This course is taught by master of machine learning who is also a gifted lecturer, who manages to clearly explain his deep understanding of the concepts. This course will make you work hard, and you need to have the right mathematics background (calculus and linear algebra). But given that, you're in for a treat!

I attended the course starting from September 2016. It is an excellent introduction to Machine Learning, covering both practical and theoretical aspects. The content is challenging and agenda of the course is tight (it is held in parallel with the live Caltech course). However, the staff is very much supportive. In particular, professor Yaser answers personally and promptly to almost all the question in the forum, making this course unique.

by
Vitaliycompleted this course, spending 16 hours a week on it and found the course difficulty to be very hard.

This is really an excellent course. It gives a real understanding of the basic concepts and methods in the world of machine learning. But this understanding is achieving through hard work, challenging tasks are available. And complexity is not an end in itself, tasks are chosen so that the solution leads to an improvement in the conceptual understanding of things. The lion's share of tasks requires setting up a computational experiment, so without good programming skills this course can become an excessive load.

The lecturer talks about the material not dispassionately, but as something very pleasant and interesting for himself. This greatly enhances the effect of perfectly prepared lectures.