Why can distributing every ranking query across many workers change training even if rows and labels are unchanged?
Classic MLLevel: AdvancedXGBoost · Learning to rank
Why can distributing every ranking query across many workers change training even if rows and labels are unchanged?
Hints
- Where are candidate pairs formed?
- Partitioning can change both pairs and normalization.
Answer
This answer is part of mentoring
Advanced questions come with a full breakdown and a mock interview with a working quant researcher. The first call is free.
Source and licence: XGBoost · Learning to rank · Apache-2.0 · answer: Quantoros editorial