method
active
method:db-mtl-dual-balancing-multi-task-learningDB-MTL (Dual-Balancing Multi-Task Learning)
The proposed method combining loss-scale balancing via logarithm transformation and gradient-magnitude balancing via maximum-norm normalization.
Neighborhood — ranked by edge-count
Concepts (4)
concept
- gradient-magnitude balancingimplementsAddressing disparity in gradient magnitudes across tasks at the gradient level
- loss-scale balancingimplementsAddressing disparity in loss magnitudes across tasks at the loss level
- Maximum gradient norm scalingimplementsScaling aggregated gradient by the maximum gradient norm among tasks.
- Exponential Moving Average (EMA)implementsUsed to estimate gradients dynamically, with forgetting rate β.
Methods (1)
method
- logarithm transformationimplementsParameter-free loss transformation applied to each task loss to equalize scales
Artifacts (1)
artifact
- The paper proposing the Dual-Balancing Multi-Task Learning method.
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.
- Novel MTL method combining loss-scale and gradient-magnitude balancing
- Core claim of the paper.
- Concise summary of the DB-MTL method from the abstract.
- Effect on gradient conflict.
- Learning paradigm that jointly learns multiple related tasks using a single model
- Training stability claim.
- Computational efficiency comparison.
- Motivation for the proposed method.