Domain-independent Dominance of Adaptive Methods
Pedro Savarese, David Mcallester, Sudarshan Babu, Michael Maire
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/lolemacs/avagradOfficialIn paperpytorch★ 0
Abstract
From a simplified analysis of adaptive methods, we derive AvaGrad, a new optimizer which outperforms SGD on vision tasks when its adaptability is properly tuned. We observe that the power of our method is partially explained by a decoupling of learning rate and adaptability, greatly simplifying hyperparameter search. In light of this observation, we demonstrate that, against conventional wisdom, Adam can also outperform SGD on vision tasks, as long as the coupling between its learning rate and adaptability is taken into account. In practice, AvaGrad matches the best results, as measured by generalization accuracy, delivered by any existing optimizer (SGD or adaptive) across image classification (CIFAR, ImageNet) and character-level language modelling (Penn Treebank) tasks.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| CIFAR-100 WRN-28-10 - 200 Epochs | AdaBound | Accuracy | 77.24 | — | Unverified |
| CIFAR-100 WRN-28-10 - 200 Epochs | AdamW | Accuracy | 79.87 | — | Unverified |
| CIFAR-100 WRN-28-10 - 200 Epochs | SGD | Accuracy | 80.95 | — | Unverified |
| CIFAR-100 WRN-28-10 - 200 Epochs | Adam (eps-adjusted) | Accuracy | 81.04 | — | Unverified |
| CIFAR-100 WRN-28-10 - 200 Epochs | AdaShift | Accuracy | 81.12 | — | Unverified |
| CIFAR-100 WRN-28-10 - 200 Epochs | AvaGrad | Accuracy | 81.24 | — | Unverified |
| CIFAR-10 WRN-28-10 - 200 Epochs | Adam (eps-adjusted) | Accuracy | 96.36 | — | Unverified |
| CIFAR-10 WRN-28-10 - 200 Epochs | AvaGrad | Accuracy | 96.2 | — | Unverified |
| CIFAR-10 WRN-28-10 - 200 Epochs | SGD | Accuracy | 96.14 | — | Unverified |
| CIFAR-10 WRN-28-10 - 200 Epochs | AdaShift | Accuracy | 95.92 | — | Unverified |
| CIFAR-10 WRN-28-10 - 200 Epochs | AdamW | Accuracy | 95.89 | — | Unverified |
| CIFAR-10 WRN-28-10 - 200 Epochs | AdaBound | Accuracy | 94.6 | — | Unverified |
| ImageNet ResNet-50 - 90 Epochs | AvaGrad | Top 1 Accuracy | 76.51 | — | Unverified |
| ImageNet ResNet-50 - 90 Epochs | AdaBound | Top 1 Accuracy | 72.01 | — | Unverified |
| ImageNet ResNet-50 - 90 Epochs | AdamW | Top 1 Accuracy | 72.9 | — | Unverified |
| ImageNet ResNet-50 - 90 Epochs | SGD | Top 1 Accuracy | 75.99 | — | Unverified |
| Penn Treebank (Character Level) 3x1000 LSTM - 500 Epochs | AdaShift | Bit per Character (BPC) | 1.27 | — | Unverified |
| Penn Treebank (Character Level) 3x1000 LSTM - 500 Epochs | AdaBound | Bit per Character (BPC) | 2.86 | — | Unverified |
| Penn Treebank (Character Level) 3x1000 LSTM - 500 Epochs | AdamW | Bit per Character (BPC) | 1.23 | — | Unverified |
| Penn Treebank (Character Level) 3x1000 LSTM - 500 Epochs | AvaGrad | Bit per Character (BPC) | 1.18 | — | Unverified |