Papers › Unified Interpretation of Softmax Cross-Entropy and Negative Sampling: With Case Study...

Unified Interpretation of Softmax Cross-Entropy and Negative Sampling: With Case Study for Knowledge Graph Embedding

14 Jun 2021ACL 2021 5arXiv:2106.07250archive 2025-07-28

Hidetaka Kamigaito, Katsuhiko Hayashi

In knowledge graph embedding, the theoretical relationship between the softmax cross-entropy and negative sampling loss functions has not been investigated. This makes it difficult to fairly compare the results of the two different loss functions. We attempted to solve this problem by using the Bregman divergence to provide a unified interpretation of the softmax cross-entropy and negative sampling loss functions. Under this interpretation, we can derive theoretical findings for fair comparison. Experimental results on the FB15k-237 and WN18RR datasets show that the theoretical findings are valid in practical settings.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

kamigaito/acl2021kge officialmentioned in papermentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Graph EmbeddingKnowledge Graph EmbeddingLink Prediction

1 archive task tag without a task page not shown.

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Link Prediction FB15k-237 RESCAL (SCE w/ LS pretrained) Hits@1 0.269 #15 of 75 Archive leaderboard report
Link Prediction FB15k-237 RESCAL (SCE w/ LS pretrained) Hits@10 0.55 #15 of 75 Archive leaderboard report
Link Prediction FB15k-237 RESCAL (SCE w/ LS pretrained) Hits@3 0.402 #15 of 75 Archive leaderboard report
Link Prediction FB15k-237 RESCAL (SCE w/ LS pretrained) MRR 0.364 #15 of 75 Archive leaderboard report
Link Prediction FB15k-237 RESCAL (SCE w/ LS) Hits@1 0.269 #16 of 75 Archive leaderboard report
Link Prediction FB15k-237 RESCAL (SCE w/ LS) Hits@10 0.548 #16 of 75 Archive leaderboard report
Link Prediction FB15k-237 RESCAL (SCE w/ LS) Hits@3 0.4 #16 of 75 Archive leaderboard report
Link Prediction FB15k-237 RESCAL (SCE w/ LS) MRR 0.363 #16 of 75 Archive leaderboard report
Link Prediction WN18RR ComplEx (SCE w/ LS pretrained) Hits@1 0.444 #49 of 75 Archive leaderboard report
Link Prediction WN18RR ComplEx (SCE w/ LS pretrained) Hits@10 0.553 #49 of 75 Archive leaderboard report
Link Prediction WN18RR ComplEx (SCE w/ LS pretrained) Hits@3 0.496 #49 of 75 Archive leaderboard report
Link Prediction WN18RR ComplEx (SCE w/ LS pretrained) MRR 0.481 #49 of 75 Archive leaderboard report
Link Prediction WN18RR ComplEx (SCE w/ LS) Hits@1 0.441 #55 of 75 Archive leaderboard report
Link Prediction WN18RR ComplEx (SCE w/ LS) Hits@10 0.546 #55 of 75 Archive leaderboard report
Link Prediction WN18RR ComplEx (SCE w/ LS) Hits@3 0.491 #55 of 75 Archive leaderboard report
Link Prediction WN18RR ComplEx (SCE w/ LS) MRR 0.477 #55 of 75 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Softmax

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections