AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Transformer-Based Learned Optimization

Erik Gärtner, Luke Metz, Mykhaylo Andriluka, C. Daniel Freeman, Cristian Sminchisescu

CVPR 2023 · Architectures and efficiency › RNNs and normalization · approach: Scalable approximation and equivariant compute · 11 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (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 changedWhereMeasuredEffect
Use Optimus instead of DiffPhy's BFGS optimizerDiffPhy human-motion reconstruction · Human3.6M validationMPJPE-G: 139.1 -> 138.6Improved
Use Optimus instead of DiffPhy's BFGS optimizerDiffPhy human-motion reconstruction · Human3.6M validationMPJPE: 82.1 -> 82.8Hurt
Use Optimus instead of DiffPhy's BFGS optimizerDiffPhy human-motion reconstruction · Human3.6M validationMPJPE-PA: 55.9 -> 57.0Hurt
Use Optimus instead of DiffPhy's BFGS optimizerDiffPhy human-motion reconstruction · Human3.6M validationFoot skate: 7.2 -> 6.5Improved
Use Optimus instead of DiffPhy's BFGS optimizerDiffPhy human-motion reconstruction · AIST dance subsetMPJPE-G: 150.2 -> 149.8Improved
Incrementally update the preconditioning matrix across iterations (ablation)Optimus · Human3.6M validationMPJPE-G: 36.5 -> 33.1Improved
Use the structured learned preconditioner rather than directly predicting the parameter update (ablation)Optimus · Human3.6M validationMPJPE-G: 37.3 -> 33.1Improved
Use the structured learned preconditioner rather than directly predicting the parameter update (ablation)Optimus · Human3.6M validationRuntime (ms): 10.7 -> 78.9Hurt
Increase optimization to a fixed 50 function evaluations without stopping early (ablation)Optimus · Human3.6M validationMPJPE-G: 33.1 -> 28.5Improved
Add BasinHopping perturbations (ablation)Optimus · Human3.6M validationMPJPE-G: 33.1 -> 25.6Improved
Increase the Adafactor MLP model size from 128x4 to 512x4Adafactor MLP · Human3.6M validationMPJPE-G: 40.6 -> 40.0Improved
Use Optimus instead of Adafactor MLPHuman-motion reconstruction optimizer · Human3.6M validationMPJPE-G: 40.0 -> 33.1Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy