AI Paper Atlas

Research map / Optimization / Optimizers

Delta-STN: Efficient Bilevel Optimization for Neural Networks using Structured Response Jacobians

Juhan Bae, Roger B Grosse

NeurIPS 2020 · Optimization › Optimizers · approach: Hypergradient bilevel optimization · 8 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
Add linearization of the hypernetwork (Centered STN to Delta-STN) (ablation)Centered STN · CIFAR10 VGG16validation loss: 0.286 -> 0.272Improved
Add linearization (Centered STN to Delta-STN) (ablation)Centered STN · CIFAR10 AlexNetvalidation loss: 0.431 -> 0.425Improved
Add linearization (Centered STN to Delta-STN) (ablation)Centered STN · CIFAR10 ResNet18validation loss: 0.222 -> 0.204Improved
Add linearization (Centered STN to Delta-STN) (ablation)Centered STN · MNISTvalidation loss: 0.041 -> 0.040Improved
Add linearization (Centered STN to Delta-STN) (ablation)Centered STN · PTBvalidation perplexity: 69.40 -> 68.63Improved
Replace STN with centered parameterization and modified training objective (ablation)STN · CIFAR10 VGG16validation loss: 0.330 -> 0.286Improved
Replace STN with centered parameterization and modified objective (ablation)STN · CIFAR10 AlexNetvalidation loss: 0.474 -> 0.431Improved
Replace STN with centered parameterization and modified objective (ablation)STN · CIFAR10 ResNet18validation loss: 0.266 -> 0.222Improved
Replace STN with centered parameterization and modified objective (ablation)STN · PTBvalidation perplexity: 70.67 -> 69.40Improved
Replace STN with centered parameterization and modified objective (ablation)STN · FMNISTtest loss: 0.218 -> 0.212Improved
Replace STN with Delta-STNSTN · MNISTtest loss: 0.041 -> 0.038Improved
Replace Bayesian optimization with Delta-STNBayesian optimization · CIFAR10 VGG16validation loss: 0.421 -> 0.272Improved

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