We propose an algorithm for meta-learning that is model-agnostic, in the sense that it is compatible with any model trained with gradient descent and applicable to a variety of different learning problems, including classification, regression, and reinforcement learning. The goal of meta-learning is to train a model on a variety of learning tasks, such that it can solve new learning tasks using only a small number of training samples. In our approach, the parameters of the model are explicitly trained such that a small number of gradient steps with a small amount of training data from a new task will produce good generalization performance on that task. In effect, our method trains the model to be easy to fine-tune. We demonstrate that this approach leads to state-of-the-art performance on two fewshot image classification benchmarks, produces good results on few-shot regression, and accelerates fine-tuning for policy gradient reinforcement learning with neural network policies.
@inproceedings{FinnAL17,
title={Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks},
author={Chelsea Finn and Pieter Abbeel and Sergey Levine},
booktitle={Proceedings of the 34th International Conference on Machine Learning},
year={2017}
}
It consists of two steps:
Step1: Base training
Step2: Meta Testing:
${WORK_DIR}/${CONFIG}/best_accuracy_mean.pth
in default.# base training
python ./tools/classification/train.py \
configs/classification/maml/cub/maml_conv4_1xb105_cub_5way-1shot.py
# meta testing
python ./tools/classification/test.py \
configs/classification/maml/cub/maml_conv4_1xb105_cub_5way-1shot.py \
work_dir/maml_conv4_1xb105_cub_5way-1shot/best_accuracy_mean.pth
Note:
Arch | Input Size | Batch Size | way | shot | mean Acc | std | ckpt | log |
---|---|---|---|---|---|---|---|---|
conv4 | 84x84 | 105 | 5 | 1 | 60.32 | 0.5 | ckpt | log |
conv4 | 84x84 | 105 | 5 | 5 | 77.03 | 0.39 | ckpt | log |
resnet12 | 84x84 | 105 | 5 | 1 | 70.44 | 0.55 | ckpt | log |
resnet12 | 84x84 | 105 | 5 | 5 | 85.5 | 0.33 | ckpt | log |
Arch | Input Size | Batch Size | way | shot | mean Acc | std | ckpt | log |
---|---|---|---|---|---|---|---|---|
conv4 | 84x84 | 105 | 5 | 1 | 46.76 | 0.42 | ckpt | log |
conv4 | 84x84 | 105 | 5 | 5 | 63.88 | 0.39 | ckpt | log |
resnet12 | 84x84 | 105 | 5 | 1 | 57.4 | 0.47 | ckpt | log |
resnet12 | 84x84 | 105 | 5 | 5 | 72.42 | 0.38 | ckpt | log |
Arch | Input Size | Batch Size | way | shot | mean Acc | std | ckpt | log |
---|---|---|---|---|---|---|---|---|
conv4 | 84x84 | 105 | 5 | 1 | 45.56 | 0.49 | ckpt | log |
conv4 | 84x84 | 105 | 5 | 5 | 60.2 | 0.43 | ckpt | log |
resnet12 | 84x84 | 105 | 5 | 1 | 57.63 | 0.53 | ckpt | log |
resnet12 | 84x84 | 105 | 5 | 5 | 72.3 | 0.43 | ckpt | log |
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。