Research map / Optimization / Optimizers
Optimizing Nondecomposable Data Dependent Regularizers via Lagrangian Reparameterization Offers Significant Performance and Efficiency Gains
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add the F1 nondecomposable regularizer with reparameterized dual ascent | DeepLabV3+ · MSCOCO Stuff 164K | mean IoU: 0.30 -> 0.32 | Helped |
| Add the F1 nondecomposable regularizer with reparameterized dual ascent | SegNet · MSCOCO Stuff, sky category | mean IoU: 77.6% -> 80% | Helped |
| Use reparameterized dual ascent with one projection every epoch instead of full projections every inner iteration | Reparameterized dual ascent · Optimization of the nondecomposable F1 objective | projection frequency: one projection operation every epoch | Helped |
Closest papers by meaning
- Learning Optimal Linear Regularizers (ICML 2019 · similarity 0.88)
- Regularized Gradient Boosting (NeurIPS 2019 · similarity 0.87)
- Functional Regularization for Representation Learning: A Unified Theoretical Perspective (NeurIPS 2020 · similarity 0.85)
- On Regularizing Rademacher Observation Losses (NeurIPS 2016 · similarity 0.85)
- Generalization Guarantees for Representation Learning via Data-Dependent Gaussian Mixture Priors (ICLR 2025 · similarity 0.84)
- A Statistical Theory of Regularization-Based Continual Learning (ICML 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.