Research map / Optimization / Optimizers
Adaptive Inertia: Disentangling the Effects of Adaptive Learning Rate and Momentum
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 |
|---|---|---|---|
| Use AdaiW instead of AdaBelief | ResNet18 · CIFAR-10 | test error: 4.97 -> 4.59 | Helped |
| Use AdaiW instead of Padam | VGG16 · CIFAR-10 | test error: 6.15 -> 5.81 | Helped |
| Use Adai instead of AdaBelief | ResNet34 · CIFAR-100 | test error: 22.44 -> 20.79 | Helped |
| Use AdaiW instead of AdaBelief | DenseNet121 · CIFAR-100 | test error: 20.49 -> 19.44 | Helped |
| Use AdaiW instead of AdaBelief | GoogLeNet · CIFAR-100 | test error: 21.15 -> 20.50 | Helped |
| Use adaptive learning rates rather than fixed learning rates | Adam dynamics · strict saddle points | saddle-point escaping | Helped |
| Use adaptive learning rates rather than fixed learning rates | Adam dynamics · flat minima selection | flat-minima preference | Hurt |
| Add momentum to SGD | SGD · strict saddle points | saddle-point escaping | Helped |
| Add momentum to SGD | SGD · flat minima selection | flat-minima preference | No clear effect |
| Use parameter-wise adaptive inertia instead of parameter-wise adaptive learning rates | Adai · loss-valley escape | flat-minima preference | Helped |
Closest papers by meaning
- Escaping Saddle Points Faster with Stochastic Momentum (ICLR 2020 · similarity 0.91)
- Adaptive Momentum and Nonlinear Damping for Neural Network Training (ICML 2026 · similarity 0.90)
- A Comprehensive Framework for Analyzing the Convergence of Adam: Bridging the Gap with SGD (ICML 2025 · similarity 0.89)
- The Marginal Value of Momentum for Small Learning Rate SGD (ICLR 2024 · similarity 0.88)
- Non-asymptotic Analysis of Biased Adaptive Stochastic Approximation (NeurIPS 2024 · similarity 0.88)
- Adaptive Methods through the Lens of SDEs: Theoretical Insights on the Role of Noise (ICLR 2025 · similarity 0.87)
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.