edit, 2nd thought: but why don't you add a link to your LI post to the text in the RecSys Part 2? tbh, imo you could also repeat the content of your linkedin post on substack and allow people to skip that part in case they have read it on LI
Sounds amazing! Looking forward to the upcoming articles in this series!
I would be very interested in learning how to choose the right recommendation approach for a given use case, what are some popular algorithms to implement, and how to balance business metrics with metrics like NDCG, MAP etc.
I will make the case that all that does not actually matter too much and instead I will be focusing on real world challenges, sorry if the focus is not going to be what you expected, but I promise it's going to be high value :)
The topic sounds promising! I would be interested into knowing also at some point how such systems are implemented from the architectural point of view.
For example, how we can ensure that an intensive run of the model can be handled asynchronously and update the recommendations!
Really interesting, excited about what will come, don’t know much about RecSys tbh
Thanks Paul means a lot! Do you think it’s easy to follow between newsletter and LI posts, or is it too advanced?
easy to follow ;)
edit, 2nd thought: but why don't you add a link to your LI post to the text in the RecSys Part 2? tbh, imo you could also repeat the content of your linkedin post on substack and allow people to skip that part in case they have read it on LI
Good points!!
Typo
"Hi Rate@K" -> "Hit Rate@K"
Thanks! At least we know it’s not AI generated ;)
so true
Sounds amazing! Looking forward to the upcoming articles in this series!
I would be very interested in learning how to choose the right recommendation approach for a given use case, what are some popular algorithms to implement, and how to balance business metrics with metrics like NDCG, MAP etc.
I will make the case that all that does not actually matter too much and instead I will be focusing on real world challenges, sorry if the focus is not going to be what you expected, but I promise it's going to be high value :)
Looking forward to it! :))
The topic sounds promising! I would be interested into knowing also at some point how such systems are implemented from the architectural point of view.
For example, how we can ensure that an intensive run of the model can be handled asynchronously and update the recommendations!
Thanks for feedback! I will be diving deep into industry papers to really show how the industry does it, so hopefully you will be satisfied by that :)