Spotify introduces randomness on the homepage to break user bubble filter and diversify recommendation #recsys2019
Bam ⚡! Slide 1, and Olivier Jeunen sets the scene! We have <user, item, timestamp>-triplets. What do we do? Efficiently compute similarity for collaborative filtering, is my hunch. A #recsys2019 paper!
Last moments of my #recsys2019
It’s been a blast! Great organization, great talks, and great workshops. Thanks ACM RecSys, see you next year in Rio!
Thread:
Broken implementations of NDCG, HR, AUC metrics were used in the #RecSys2019 paper 'Are we really making much progress? A worrying analysis of recent neural recommendation approaches' Dacrema et. al.
dl.acm.org/doi/abs/10.114…
This covers a more fundamental issue. (1/3)
The BBC journey for deploying a Recommender System that takes into account their editorial policies as well as the legal framework (eg GDPR) #recsys2019
Be mindful of using metrics that measure something similar to what your algorithm optimises -illustration through various diversification approaches and metrics #recsys2019
Visit our poster ACM RecSys on “Efficiënt similarity computation for collaboratie filtering in dynamic environments” or the presentation in the plenary session up next, by Olivier Jeunen. #recsys2019
Presenting at #recsys2019 about content #recommendation Chegg. It was a great week of exchanging knowledge, learning industry applications and meeting fellow #AI and #MachineLearning practitioners. Looking forward to implementing some of the new ideas soon. #Data
Is this (finally!) one of first mainstream uses of an user-configurable intent for recommender engine? Cc #recsys2019
theverge.com/2019/10/1/2089…