claim
active
claim:the-logarithm-transformation-is-simpler-and-more-effective-than-imtl-l-because-it-is-parameter-freeThe logarithm transformation is simpler and more effective than IMTL-L because it is parameter-free.
Comparison of loss-scale balancing techniques.
Source paper
extracted_from(2023) · Baijiong Lin · Weisen Jiang · Feiyang Ye · Yu Zhang +5
Neighborhood — ranked by edge-count
Findings (1)
finding
- Comparison of loss-scale balancing with IMTL-L.
Communities (3)
community
- Dual-balancing multi-task learningmembers_ofDB-MTL jointly balances loss scale and gradient magnitude, benchmarked on NYUv2 and Office-31.
- Dual balancing multi-task learningmembers_ofDB-MTL combines loss-scale and gradient-magnitude balancing, benchmarked across NYUv2, Cityscapes, QM9, and Office datasets.
- Parameter-free logarithm transformation for multi-task learning that improves gradient balancing methods like PCGrad and Nash-MTL across vision benchmarks.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Mathematical relationship between IMTL-L and log transformation.
- Compared to IMTL-L: parameter-free, no extra computational cost, achieves same theoretical goal
- We find that the logarithm transformation also benefits existing gradient balancing methods.quote0.810Key finding showing the broader utility of the log transformation.
- Generalization of the loss transformation.
- Effectiveness of logarithm transformation as a plug-in for gradient balancing methods.
- Parameter-free loss transformation applied to each task loss to equalize scales
- Prior loss-balancing method using learnable loss transformation; logarithm approach recovers this
- C-Linda code is easier to understand than the Parlog86 version [for the client-server problem].claim0.733Subjective but argued comparison.