Research map / Optimization / Optimizers
Augment Your Batch: Improving Generalization Through Instance Repetition
Check 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 |
|---|---|---|---|
| Add batch augmentation with M=4 | ResNet50 · ImageNet | Top-1 validation accuracy: 76.3% to 76.86% | Helped |
| Add batch augmentation with M=4 | Transformer base · WMT En-De | BLEU: 26.88 BLEU -> 27.13 BLEU | Helped |
| Add batch augmentation with M=10 | Word-level LSTM · Penn Treebank | test perplexity: 58.8 ppl -> 58.6 ppl | Helped |
| Add batch augmentation with M=4 | ResNet50 · ImageNet, mixup α=0.2 | Top-1 validation accuracy: 76.70% -> 77.04% | Helped |
| Add batch augmentation with M=10 | ResNet44 · CIFAR-10, mixup α=0.2 | Top-1 validation accuracy: 94.60% -> 95.55% | Helped |
| Increase the number of batch augmentations per sample (ablation) | ResNet44 with Cutout · CIFAR-10 | validation classification error | Helped |
| Add batch augmentation | ResNet50 · ImageNet | validation accuracy: 76.3% -> 76.86% | Helped |
| Use 10 instead of 4 augmentations per sample (ablation) | Distributed ResNet50 · ImageNet | validation error | Helped |
| Increase local batch size from 64 to 128 (ablation) | ResNet50 · ImageNet | throughput: 234.4 -> 247.9 | Helped |
| Use batch augmentation rather than standard training | ResNet44 · CIFAR-10, same shortened training regime | validation accuracy: 93.07% -> 94.15% | Helped |
Closest papers by meaning
- On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima (ICLR 2017 · similarity 0.91)
- Train longer, generalize better: closing the generalization gap in large batch training of neural networks (NeurIPS 2017 · similarity 0.90)
- Four Things Everyone Should Know to Improve Batch Normalization (ICLR 2020 · similarity 0.89)
- Tradeoffs in Data Augmentation: An Empirical Study (ICLR 2021 · similarity 0.89)
- Deconstructing the Regularization of BatchNorm (ICLR 2021 · similarity 0.89)
- Small Batch Size Training for Language Models: When Vanilla SGD Works, and Why Gradient Accumulation is Wasteful (NeurIPS 2025 · similarity 0.88)
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.