The cost of an item-based recommender can depend a lot on the implementation. If you make the access to all of the related items of a single item be O(1), then the cost of item-based recommendations is only proportional to the size of the user history that you consider.
Basically what happens here is that much of the work of the user based recommendation system has been move to an off-line process. This very much limits your update rate, but related items don't change terribly quickly. On Sat, Feb 27, 2010 at 10:28 AM, Claudio Martella < [email protected]> wrote: > "They do have notably different properties. For instance, the running > time of an item-based > recommender scales up as the number of items increases, whereas a > user-based recommender’s > running time goes up as the number of users increases. > -- Ted Dunning, CTO DeepDyve
