Recently we’ve discussed a lot about recommendation systems. Today we want to show you some of the efforts made to find solutions to their known drawbacks, such as the determination of similar users, items or batch-processed recommendations.
An obvious improvement to recommender systems is to try and get recommendations in real time. Here, the author shows how that can be attained : using graph-based input for recommendations.
The main problem of the real time approach is complexity, so he goes on explaining how graph-based recommenders do not suffer from it. However this example is still quite simple and we are not shown any performance metrics scaled to current situations.
What do you think ? Are graphs the solution or is it just another buzzword ?
