Introduction to Recommender Systems

https://www.coursera.org/course/recsys

by swGooF • 13 years ago

194 48 13 years ago

ghc2• 13 years ago

I can't judge the coursera course, but for anyone who is interested in this field and wants a gentle introduction, I high recommend Programming Collective Intelligence (http://shop.oreilly.com/product/9780596529321.do). It covers many of the types of recommender systems that the coursera course is likely to cover, and comes with a lot of nice Python code examples.

It's highly useful knowledge too. I ran across so many startups that needed recommender systems that I launched a company called Algorithmic.ly (http://algorithmic.ly) to help companies without the expertise integrate recommendation systems and other types of algorithms into their projects.

riggins2• 13 years ago

Thanks for the recommendation.

phektus• 13 years ago

Meta. Sorry :P

dougk7• 13 years ago

PCI is a great book. One book I'll recommend after reading it is "Mining of Massive Datasets" (http://infolab.stanford.edu/~ullman/mmds.html) - Chapter 9 especially. It goes into more depth than what's covered in PCI and it's not too difficult to follow.

I'm working on a personal project in recommender systems so I look forward to enjoy this course.

kops2• 13 years ago

Thanks for that link. Chapter 7 helped answer some of the questions I had after reading PCI.

arms2• 13 years ago

I was about to buy PCI a few days ago, but was worried that the information may be out of date, considering it was published six years ago. Do you think most of the material is still pertinent?

ghc2• 13 years ago

I believe all of the material is still pertinent. It primarily uses the MovieLens data set and mostly builds off of textbook techniques and generic Python. It's not a book that will go out of data quickly.

kops2• 13 years ago

Go for it. It's a great book and except for references to friendster as the big social network and facebook a newbie, nothing else seemed dated to me. Most of the APIs used in the book are still pretty much available except probably(??) delicious API.

clinth• 13 years ago

I was a technical reviewer of that book (although no knowledge of PCI in practice), and I found it to be less moment-specific and more general grounding and strategy than I expected. It really is a good book.

arms2• 13 years ago

Thanks for the recommendations, all. Just got my copy delivered yesterday and I'm looking forward to digging into it.

tungwaiyip• 13 years ago

In this short article I have implemented the algorithm in the Programming Collective Intelligence in 9 lines of code

http://tungwaiyip.info/2012/Collaborative%20Filtering.html

eldog_• 13 years ago

I'd be interested in knowing how much deep learning is changing the algorithms used in this field, given the performance of restricted boltzmann machines on the netflix data set http://www.cs.utoronto.ca/~hinton/absps/netflixICML.pdf.

colincsl2• 13 years ago

I am surprised that there is no mention of this in the course syllabus -- in fact it looks like a lot of recent techniques that are missing. They don't even talking about LSA(/SVD)-based methods until the end of the course.

jere• 13 years ago

Dumb, general question: should it really be a surprise that an Introduction course doesn't use all the recent techniques?

elq• 13 years ago

You'd be surprised how overwhelmingly common and effective very simple methods like logistic regression and basic decision trees are for such systems.

Further, RBMs and other deep learning tools require a significantly more sophisticated mathematical background than algebra and a much broader understanding overall.

Irishsteve• 13 years ago

The netflix prize touched on one of many areas related to recommender systems.

As mentioned already, very simple methods can be really effective. Things such as the UI are also known to have a big impact on how 'useful' people find the recs.

msellout• 13 years ago

Not much yet. It's not well enough understood yet.

riggins2• 13 years ago

coursera is killing me with courses I want to take.

DannoHung3• 13 years ago

I just wish you could take the classes at your own pace. I'm getting very little out of the "certifications" or whatever. And the TA's/Discussion forums tend towards pretty useless because of how disorganized they are.

srom• 13 years ago

Download the resources and you'll be able to take the class at your own pace. There are several tools for that purpose on github. e.g this one: https://github.com/dgorissen/coursera-dl

As a special gift, it's allow you to archive the courses for later review.

fantnn• 13 years ago

Agree 100%, it's maddening. A lot of the people that would get the most out of these courses don't have time to adhere to a strict schedule like they expect you too. I wonder if it has to do with colleges realizing just how disruptive it could be to there business model.

dorkrawk• 13 years ago

I've done this with Coursera classes. I just sign up for them so that they're always saved in "My Courses" and go through the video lectures at my leisure. I miss the grading and a lot of the live discussion, but the notes and lectures are certainly valuable.

michaelrbock• 13 years ago

Being able to go at your own pace (and course content quality) is why I think Udacity is a better course provider than Coursera (even though Udacity is clearly loosing in market share).

wcbeard10• 13 years ago

I believe their Discrete optimization course is at least somewhat self-paced. [1] Also, not sure if all of the Stanford/edX courses are, but Medical Stats is, and it's been really helpful for keeping my attention.

[1] https://www.coursera.org/course/optimization [2] medstats.class.stanford.edu

victorhooi• 13 years ago

Hmm, seems interesting = I'm currently doing the Machine Learning one also via Coursera, run by Andrew Ng, and it's good gentle introduction to the subject.

It's a shame we can't view the course content for this one earlier...haha.

pallandt3• 13 years ago

I may be wrong, but I think it's the 1st time the course is being offered, I don't personally remember seeing it before in their course list.

cgag• 13 years ago

I think that was one of the first courses they offered actually.

mark_l_watson• 13 years ago

If you want a quick start without taking a class, install the Apache Mahout project - one of the Hadoop map-reduce examples is a recommendation system. You can hack away, and run on Elastic MapReduce if you need to scale. (https://cwiki.apache.org/confluence/display/MAHOUT/Recommend...)

axansh• 13 years ago

Thanks coursera.

Thank you very much.

Looking forward for this course.

You save my some $$$ :) Will surely donate you.

oneeyedpigeon2• 13 years ago

I couldn't possibly recommend a site that requires javascript to display any content whatsoever.

jwr• 13 years ago

Why not? JavaScript has become a part of the web, just as much as HTML and CSS. You might as well shun sites that use the <ul> tag, or CSS to style content.

oneeyedpigeon2• 13 years ago

JavaScript should certainly be used to enhance content and ad functionality. But the core content should always be available without JS, since not all consumers of that content will have JS enabled, or be able to interpret it.

Ntrails• 13 years ago

<div style="text-align:center; margin-bottom:10px;">Please use a <a href="/browsers">modern browser </a> with JavaScript enabled to use Coursera.</div>

Good job there folks

rypskar• 13 years ago

At least they use noscript and give you a warning if you don’t have javascript enabled, there are so many sites that don’t check if javascript and cookies are enabled

arek2• 13 years ago

I'll drop a link to my (non-free) e-book: http://arek-paterek.com/book/

javindo• 13 years ago

Interesting anecdote: A graduate of my university works for Google who originally had a very complex "machine learning pipeline" for the product recommendations but he has since re-implemented the feature in, as he calls it, a "much simpler bloom filter algorithm".

fmela• 13 years ago

So many great online courses, and so little time!

Introduction to Recommender Systems

Related Stories