Research map / Optimization / Optimizers
Adam Can Converge Without Any Modification On Update Rules
Yushun Zhang, Congliang Chen, Naichen Shi, Ruoyu Sun, Zhi-Quan Luo
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use small β₂ in the paper's divergent hyperparameter region | Adam on a fixed finite-sum function class · Constructed smooth convex function | iterates and function values: null | Hurt |
| Increase β₂ | Adam on the constructed function, initialized at x = -5 · Constructed function | gradient norm: β₂ = 0, 0.1,0.3,0.5,0.7,0.9,0.99 | Improved |
| Use small β₁ and β₂ | Adam · MNIST and CIFAR-10 | optimization error: 1.4 times | Hurt |
| Use small β₁ and β₂ | Adam · Convex function (defined in the paper) | optimization error: 70 times | Hurt |
| Increase β₂ | Adam with D₀ > 0 · Stochastic quadratic function | size of convergence region: null | Improved |
| Reduce batchsize, which increases the number of batches n | Adam with β₁ fixed at 0.9 · MNIST | training loss: null | Hurt |
| Increase n | Adam on the constructed function · Function initialized at x = 1 | optimality gap: n =5, 10 ,15, 20 | Mixed |
Closest papers by meaning
- ADOPT: Modified Adam Can Converge with Any $\beta_2$ with the Optimal Rate (NeurIPS 2024 · similarity 0.89)
- A Comprehensive Framework for Analyzing the Convergence of Adam: Bridging the Gap with SGD (ICML 2025 · similarity 0.89)
- Adam under Generalized Smoothness with Second-Moment-Type Stochastic Gradients (NeurIPS 2026 · similarity 0.89)
- Adam Exploits $\ell_\infty$-geometry of Loss Landscape via Coordinate-wise Adaptivity (ICLR 2025 · similarity 0.86)
- The Hidden Ratio in Adam: Stable Structure, Compression, and Sign Dynamics (NeurIPS 2026 · similarity 0.86)
- On Convergence of Adam for Stochastic Optimization under Relaxed Assumptions (NeurIPS 2024 · similarity 0.86)
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.
