For sampled-softmax recommender training under a fixed memory budget B=n*k, the paper recommends the largest feasible batch size and fewest negatives, n = k = sqrt(B).
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Batch Size or Negatives? A Selection Rule for Memory-Constrained Recommender Training
For sampled-softmax recommender training under a fixed memory budget B=n*k, the paper recommends the largest feasible batch size and fewest negatives, n = k = sqrt(B).