%0 Generic %1 Ying2022 %A Zhao, Ying %A Tian, Zhiliang %A Yao, Huaxiu %A Zheng, Yinhe %A Lee, Dongkyu %A Song, Yiping %A Sun, Jian %A Zhang, N. %D 2022 %K posted_with_chatgpt machine-learning NLP meta-learning low-resource-nlp to-check %P 583-595 %R 10.48550/arXiv.2203.11670 %T Improving Meta-learning for Low-resource Text Classification and Generation via Memory Imitation %U https://www.semanticscholar.org/paper/998d8b600b3f6bf20c76c217dff4f6c4ff6f95c5 %X Building models of natural language processing (NLP) is challenging in low-resource scenarios where limited data are available. Optimization-based meta-learning algorithms achieve promising results in low-resource scenarios by adapting a well-generalized model initialization to handle new tasks. Nonetheless, these approaches suffer from the memorization overfitting issue, where the model tends to memorize the meta-training tasks while ignoring support sets when adapting to new tasks. To address this issue, we propose a memory imitation meta-learning (MemIML) method that enhances the model’s reliance on support sets for task adaptation. Specifically, we introduce a task-specific memory module to store support set information and construct an imitation module to force query sets to imitate the behaviors of support sets stored in the memory. A theoretical analysis is provided to prove the effectiveness of our method, and empirical results also demonstrate that our method outperforms competitive baselines on both text classification and generation tasks.

@JournalArticle{Ying2022, abstract = {Building models of natural language processing (NLP) is challenging in low-resource scenarios where limited data are available. Optimization-based meta-learning algorithms achieve promising results in low-resource scenarios by adapting a well-generalized model initialization to handle new tasks. Nonetheless, these approaches suffer from the memorization overfitting issue, where the model tends to memorize the meta-training tasks while ignoring support sets when adapting to new tasks. To address this issue, we propose a memory imitation meta-learning (MemIML) method that enhances the model’s reliance on support sets for task adaptation. Specifically, we introduce a task-specific memory module to store support set information and construct an imitation module to force query sets to imitate the behaviors of support sets stored in the memory. A theoretical analysis is provided to prove the effectiveness of our method, and empirical results also demonstrate that our method outperforms competitive baselines on both text classification and generation tasks.}, added-at = {2023-11-09T09:47:35.000+0100}, author = {Zhao, Ying and Tian, Zhiliang and Yao, Huaxiu and Zheng, Yinhe and Lee, Dongkyu and Song, Yiping and Sun, Jian and Zhang, N.}, biburl = {https://www.bibsonomy.org/bibtex/2886de95b564913827df5011d9feca49f/tomvoelker}, day = 22, description = {This paper introduces a novel memory imitation meta-learning approach aimed at improving task adaptation for low-resource text classification and generation. It leverages a task-specific memory module to better utilize support sets, thereby enhancing the model's adaptability to new tasks.}, doi = {10.48550/arXiv.2203.11670}, interhash = {16ca1f5302b17291863f68265a0b79c2}, intrahash = {886de95b564913827df5011d9feca49f}, keywords = {posted_with_chatgpt machine-learning NLP meta-learning low-resource-nlp to-check}, month = {3}, pages = {583-595}, timestamp = {2023-11-09T09:47:35.000+0100}, title = {Improving Meta-learning for Low-resource Text Classification and Generation via Memory Imitation}, url = {https://www.semanticscholar.org/paper/998d8b600b3f6bf20c76c217dff4f6c4ff6f95c5}, year = 2022 }