A Recommender System is a process that seeks to predict user preferences. This Specialization covers all the fundamental techniques in recommender systems, from non-personalized and project-association recommenders through content-based and collaborative filtering techniques, as well as advanced topics like matrix factorization, hybrid machine learning methods for recommender systems, and dimension reduction techniques for the user-product preference space. This Specialization is designed to serve both the data mining expert who would want to implement techniques like collaborative filtering in their job, as well as the data literate marketing professional, who would want to gain more familiarity with these topics. The courses offer interactive, spreadsheet-based exercises to master different algorithms, along with an honors track where you can go into greater depth using the LensKit open source toolkit. By the end of this Specialization, you’ll be able to implement as well as evaluate recommender systems. The Capstone Project brings together the course material with a realistic recommender design and analysis project.
LP
Interesting course, good overview, and presentation of the topic to those who are not familiar with RS.Could have been 5 stars if the "developer" modules were available on Python. That's a big fail.
LL
a great class, I learned some insight in these algorithms
MB
It was a great course! Everyone from variety of backgrounds like MS/PhD students or industry professionals that has basic Information Retrieval and ML knowledge could understand the course content.
DD
Very good. Per closing comments, it probably needs an update (since 2016) as this is active, progressive area.
TL
I think I am on the right track to changing my career from java engineer from data scientist, this course is one of the best start point
AD
Loved it...many thanks Prof. Joe for bringing this content to Coursera
LL
wonderful!!! They teach a lot what I did not expect!
LL
great courses! They invite a lot of interviews to let me understand the sea of recommend system!
LL
Please update the specialization, it's 2022, and the course slides are from 2016.
SS
Thank you so very much to open my eye see more view of recommendation field not only algorithms but use case and many trouble-shooting in worldwide business, moreover interview with noble professor.
CS
A lot of very in detail theories and metrics. I wish it could have more hands on experience.
AG
Programming Assignments are not clear enough and the quiz for the last one seems to be a bit off.
Showing: 20 of 141
One of the best courses I have taken on Coursera. Choosing Java for the lab exercises makes them inaccessible for many data scientists. Consider providing a Python version.
well one thing I am struggling with programming in JAVA. Would not it be handy to have option to do assignment using languages like python/R? which are basically language of choice for data scientists and also easy to have grasp on for newbies. one more thing some time I just get stuck and felt like now way out. I did not get any answer/help form posts on the forum .
Very good course, Very well delivered and paced. Could benefit from an update?
(1) No Python honors option? (e.g. per LensKit platform used has moved to Python since 2018)
....this is why I did not take up the honors option
(2) Deep Learning based RS?
(3) 2013 RS paper cited not accessible and newer RecSys papers since?
Honors track should be in Python. The subjective questions of the evaluation lacks clarity in some cases.
- Too slow, too much wandering around instead of focusing on the concepts - Outdated coding exercises that don't integrate properly with modern IDEs - Too much emphasis on spreadsheets, too little emphasis on coding - The coding exercises made in Java put too much emphasis on unrelated stuff, not just because of the language, but because of how they are prepared.
Coding assignment should not be just restricted to java
Too basic and too repetitive (the videos could be half as long)
The now old trick of quite many data courses on Coursera of lulling it through the instructional material and only raising the bar when it comes to the quizzes is abundant in this course.
If you are interested in 25 minute -videos (??!) showing how the instructor chooses a lightbulb on Amazon then you will surely like this course. If not, there's better elsewhere.
There is no math in this course and it does not use Python. Therefore this course does a terrible job of preparing you for interview questions on Recommender systems. Personally I thought this course was a waste of my time and money. However the final excel exercise actually had some useful information, but it was only a 10 minute exercise after many hours of useless lectures. I could have done the same exercise for free.
Fantastic course. Lecturers have extensive experience in this field. Lectures include interviews with people who have successfully implemented recommender systems in their products or who are researching the permutations, challenges and extensions to recommender system development. Not only does the course provide the chance to build your own recommender systems (optional) but also highlights the complexities and opportunities for refining and improving recommendations. I highly recommend this course to anyone building recommendation systems.
Exceptional quality.The course content is comprehensive and practical enough applied at workplaces.
Guest lectures are super helpful and assignments are very practical yet make you think.
Thank you Coursera and Minnesota professors for this amazing course and wonderful opportunity for people like me with no background in recommendation systems learn the best research methods and practices in this field.
Nice introduction to recommender systems for those who have never heard about it before. No complex mathematical formula (which can also be seen by some as a downside).
I am confused using Java for programming, it is better using python or R in the next course
Big fan of Prof. Konstan ! Amazing how he subtly puts a joke or two in middle of explaining something. I feel the course could benefit from a little more programming content / assignments, and some more math behind the ideas presented. It presents really good philosophy behind recommendation systems and definitely got me hooked into this amazing field. However the course itself is very diluted and could use some math.
Interesting course, good overview, and presentation of the topic to those who are not familiar with RS.
Could have been 5 stars if the "developer" modules were available on Python. That's a big fail.
Would be nice to complete the honours track using python.
The pace is too slow. Lectures spend lots of time on examples, and all kinds of possible variables.
These make stories very long, and badly-structured. It may be better to introduce only one concept at any moment, and discuss the problem and the solution immediately after mentioning the concept. That will help students to focus on the point and get it right sooner. It's good to combine all these concepts together after we've known everything, but not at the very beginning.
Also the programming assignment is really bad. As a CS student, I spent almost 90% of time on realizing the architecture, tools and libraries. I don't think these third-party libraries are helpful here. The same tasks can be implemented by pure Java code even more efficiently (for coding). Most non-CS students will find it difficult to use, while CS students can learn only little from the assignment since the core ideas to implement are far too easy.
I can feel how much knowledge lectures expect us to get from this lecture, but it really needs a rebuilding. Maybe trying to put a self limitation on video length will be a good start. Expressing a brief idea in a short video, and allowing students to consume one video even with only a piece of time, should be one of the most appealing part in flip-classroom.
Pros:
Some useful terminology if you want to ever communicate with someone who does recommender systems.
Cons:
Very diluted content.
Mostly large text slides with the presenter talking in a monotone voice.
Programming exercises are done in Java and require deploying an IDE + an unused open source project developed by the authors. Hint to the authors: use Python, R or Octave like everyone does.
Some of the questionaries are ambiguous.
The content of this course is solid. It's a good introduction to content based and non-personailzed recommender systems. However, the presentation is poor. The course is largely based around videos which appear to be single takes. Snappier, well edited videos would have been better and, as a result, I often found myself skimming the transcripts rather than watching the videos.
As an introductory course, the content was good. But I wish the approach was more analytical and more hands on. Rather than history of Recommender systems & what happened in the 90s, I would have been happier if the course was able to throw light on the latest stuff in this field, the latest mathematical techniques etc.