Practical Learning-to-Rank: Deep, Fast, Precise - Roman Grebennikov
DataTalksClub ⬛
13,690 views
Links:
Slides: https://metarank.github.io/datatalks-...
Metarank: https://github.com/metarank/metarank
MSRD dataset: https://github.com/metarank/msrd
Free data engineering course: https://github.com/DataTalksClub/data...
Join DataTalks.Club: https://datatalks.club/slack.html
Our events: https://datatalks.club/events.html
9 Comments
@pranayroy 2 years ago
Appreciate this talk!!
@ritwikpm 5 months ago (edited)
Great talk! But is ndcg the right metric here...?
Consider two sessions. Session 1: user purchases from position 1, session ends. NDCG = 1. Session 2: user clicks 10 items, purchases nothing, session ends. All grades are [1,1,1,1,1,1,1,1,1,1], IDCG is the same, NDCG = 1. Both sessions score identically despite one being a perfect outcome and the other being a complete failure. NDCG cannot distinguish between the two....
Consider two sessions. Session 1: user purchases from position 1, session ends. NDCG = 1. Session 2: user clicks 10 items, purchases nothing, session ends. All grades are [1,1,1,1,1,1,1,1,1,1], IDCG is the same, NDCG = 1. Both sessions score identically despite one being a perfect outcome and the other being a complete failure. NDCG cannot distinguish between the two....
Practical Learning-to-Rank: Deep, Fast, Precise - Roman Grebennikov
DataTalksClub ⬛
13,690 views
Links:
Slides: https://metarank.github.io/datatalks-...
Metarank: https://github.com/metarank/metarank
MSRD dataset: https://github.com/metarank/msrd
Free data engineering course: https://github.com/DataTalksClub/data...
Join DataTalks.Club: https://datatalks.club/slack.html
Our events: https://datatalks.club/events.html