Looks like a promising beginning to a recommendation-engine startup.
It's good to see pure machine-learning based startups get going. I'm looking forward to a lot of interesting stuff happening and massive growth in this area over the next few years. Will be fun to be able to shove a load of data through a variety of ML system's APIs at will.
I've got kind of a love-hate relationship with Programming Collective Intelligence.
In many ways it's an awesome book. It covers a lot of territory with relative grace and in clear language. On the other hand, it's often simplified to the point that the versions of things covered in there aren't really suitable for more than toy applications.
My fear is that it brings the low end of some of these fields in reach of folks who aren't used to working with the sort of material that they'll need to get up to the next notch -- which is really where the practical applications begin. I suspect that could be frustrating for some.
Unfortunately it'd be hard to make sense of a reading list that I'd put together since unlike academic research where you tend to bore down deeper and deeper along a certain path, I've been picking up ideas in more of a grab-bag fashion. Some of the papers that I like aren't especially good papers, but happened to be the connector between two ideas that I'd been kicking around.
One of the better general introductions, with a lot of good references at the end is:
That does take a significantly more traditional approach to recommendations algorithms than we do and also the algorithms there tend to do best in cases where (not surprising given the context) the number of users is much larger than the number of items being rated.
It's good to see pure machine-learning based startups get going. I'm looking forward to a lot of interesting stuff happening and massive growth in this area over the next few years. Will be fun to be able to shove a load of data through a variety of ML system's APIs at will.