[HN Gopher] Cosine Similarity Isn't the Silver Bullet We Thought...
___________________________________________________________________
Cosine Similarity Isn't the Silver Bullet We Thought It Was
Author : skelts
Score : 8 points
Date : 2025-01-13 16:12 UTC (3 days ago)
(HTM) web link (www.shaped.ai)
(TXT) w3m dump (www.shaped.ai)
| skelts wrote:
| Netflix and Cornell University researchers have exposed
| significant flaws in cosine similarity. Their study reveals that
| regularization in linear matrix factorization models introduces
| arbitrary scaling, leading to unreliable or meaningless cosine
| similarity results. These issues stem from the flexibility of
| embedding rescaling, affecting downstream tasks like
| recommendation systems.
| throwaway314155 wrote:
| > Netflix and Cornell University researchers
|
| Strange times.
| emschwartz wrote:
| Very interesting. Thanks for sharing!
___________________________________________________________________
(page generated 2025-01-16 23:00 UTC)