Countering Mainstream Bias via End-to-End Adaptive Local Learning

1 code implementation • 13 Apr 2024 • Jinhao Pan, Ziwei Zhu, Jianling Wang, Allen Lin, James Caverlee

In this paper, we identify two root causes of this mainstream bias: (i) discrepancy modeling, whereby CF algorithms focus on modeling mainstream users while neglecting niche users with unique preferences; and (ii) unsynchronized learning, where niche users require more training epochs than mainstream users to reach peak performance.

Collaborative Filtering