Research map / Optimization / Optimizers
Towards Practical Adam: Non-Convexity, Convergence Theory, and Mini-Batch Acceleration
Congliang Chen, Li Shen, Fangyu Zou, Wei Liu
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 |
|---|---|---|---|
| increase exponent r of theta_t (r=0 to 1) in Generic Adam (ablation) | Generic Adam · 1-D counterexample | convergence | Improved |
| use theta_t increasing to 1 (r>0) instead of constant theta (r=0, original Adam) (ablation) | Generic Adam · LeNet on MNIST and ResNet-18 on CIFAR-100 | training loss and test accuracy | Improved |
| increase batch size for mini-batch SGD and SGD-momentum (ablation) | SGD · Quadratic minimization | optimal learning rate: 0.005 -> 0.04 | Mixed |
| increase batch size for mini-batch Adam (ablation) | Adam · Quadratic minimization | optimal learning rate: between 0.02 to 0.04 | No clear effect |
| increase batch size (32, 64, 128) for mini-batch Adam (ablation) | Adam · MNIST and CIFAR-100 | training loss | Improved |
| increase batch size (32, 64, 128) for mini-batch Adam (ablation) | Adam · MNIST and CIFAR-100 | test accuracy and test loss | No clear effect |
| use adaptive methods (Adam, AMSGrad) instead of SGD-M | Transformer XL · WikiText-103 | training loss | Improved |
Closest papers by meaning
- A Comprehensive Framework for Analyzing the Convergence of Adam: Bridging the Gap with SGD (ICML 2025 · similarity 0.93)
- Understanding the Generalization of Stochastic Gradient Adam in Learning Neural Networks (NeurIPS 2025 · similarity 0.91)
- Adam-mini: Use Fewer Learning Rates To Gain More (ICLR 2025 · similarity 0.90)
- Adam with model exponential moving average is effective for nonconvex optimization (NeurIPS 2024 · similarity 0.89)
- The Hidden Ratio in Adam: Stable Structure, Compression, and Sign Dynamics (NeurIPS 2026 · similarity 0.89)
- Adam Reduces a Unique Form of Sharpness: Theoretical Insights Near the Minimizer Manifold (NeurIPS 2025 · similarity 0.89)
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.
