Five Recommendation Algorithms No Recommendation Engine Is Whole Without

Post by drewcoo » Tue, Nov 22, 2022, 5:29 PM UTC

carschno wrote: post #2

I feel like these click-baity titles have finally reached a state in which they work inversely for me. I have recently noticed that I would never click on something that promises "X things you must know" because I have finally internalized that the article behind will be superficial and useless at best. The impulse of "let's at least skim it in case it includes something useful", that used to be triggered by the clic…

The enumerated listicle, you mean? Those seem so oldie oughties! Surprising they still work.

I skipped the article and came here for the comments in case there was anything useful to learn.

Post by nkrvavica » Wed, Nov 23, 2022, 9:27 AM UTC

charcircuit wrote: post #8

>The most powerful recommendation algorithms are made especially for graph data This isn't true. The most powerful recommendation systems don't use graph algorithms. What recommendation system is using breadth first search over vector search for doing candidate generation?

Hey! I read many papers that say that graph algorithms and graph neural networks have very good results in this area. In this article I explained just few of them and gave some examples how to use them for this use case. Of course if you combine different algorithms and techniques you might get better results depending on what you want to achieve.

Re: Five Recommendation Algorithms No Recommendation Engine Is Whole Without

Re: Five Recommendation Algorithms No Recommendation Engine Is Whole Without