%0 Generic %1 Chelsea2017 %A Finn, Chelsea %A Abbeel, P. %A Levine, S. %D 2017 %K meta-learning deep-learning fast-adaptation low-resource-nlp to-check posted_with_chatgpt %P 1126-1135 %T Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks %U https://www.semanticscholar.org/paper/c889d6f98e6d79b89c3a6adf8a921f88fa6ba518 %X 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 few-shot image classification benchmarks, produces good results on few-shot regression, and accelerates fine-tuning for policy gradient reinforcement learning with neural network policies.

@JournalArticle{Chelsea2017, abstract = {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 few-shot image classification benchmarks, produces good results on few-shot regression, and accelerates fine-tuning for policy gradient reinforcement learning with neural network policies.}, added-at = {2023-11-09T09:52:59.000+0100}, author = {Finn, Chelsea and Abbeel, P. and Levine, S.}, biburl = {https://www.bibsonomy.org/bibtex/22358a45730de03c7843fd68be0798ad0/tomvoelker}, day = 9, description = {This seminal work by Finn et al. introduces the concept of model-agnostic meta-learning (MAML), a method that enables rapid adaptation to new tasks, which is particularly beneficial for low-resource settings in NLP.}, interhash = {52c1ed07488c96d0a0d4f19d1c4d635f}, intrahash = {2358a45730de03c7843fd68be0798ad0}, keywords = {meta-learning deep-learning fast-adaptation low-resource-nlp to-check posted_with_chatgpt}, month = {3}, pages = {1126-1135}, timestamp = {2023-11-09T09:52:59.000+0100}, title = {Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks}, url = {https://www.semanticscholar.org/paper/c889d6f98e6d79b89c3a6adf8a921f88fa6ba518}, year = 2017 }