Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
zhangy03 ae9bda04d0 | 2 years ago | |
---|---|---|
docs | 2 years ago | |
related_paper | 2 years ago | |
README.md | 2 years ago | |
few-shot learning0607.pptx | 2 years ago |
小样本学习(few-shot learning):在人类快速学习能力的启发下,研究人员希望机器学习模型在学习了一定类别的大量数据后,对于新的类别,只需要少量的样本就能快速学习,这就是Few-shot Learning要解决的问题。
举例:摘要:以历史文字、图像和口述史资料的分析为切入点,从渊源和修复两个方面,扼要探讨了桑珠孜宗堡的变迁真相及复原再现的特殊意义.
关键词:"口述史,桑珠孜宗堡,修复"
label: "是"
P1: "摘要:"+text+",关键词:"+keyword+label+"真实关键词\n\n"
其中label_map = {"0": "不是", "1": "是"}
P2: "摘要:" + text + ",关键词:" + keyword +label+"摘要提取的关键词\n\n"
label_map = {"0": "不是", "1": "是"}
P3: “提取的关键词对吗?”+label+"正确。"+“摘要:”+text+“关键词:”+keyword +"\n\n"
label_map = {"0": "不", "1": "很"}
P4: "摘要:"+text +",关键词:"+keyword +",这是"+label + "准确的\n\n"
label_map = {"0": "不", "1": "狠"}
P5: "摘要:"+text+",关键词:"+keyword+label+"准确的关键词\n\n"
其中label_map = {"0": "不是", "1": "是"}
PPT包含以下几个部分:
(1)论文《True Few-Shot Learning with Language Models》概述,给以后的少样本学习指明了方向,用实验验证了模型选择的困难。
(2)智源大会分享的FewNLU小样本框架:主要包括规范了少样本学习的评测方法和一整套实验的框架,减少overfitting带来的不真实的结果,使得评测是科学的,有意义的。提出了基于相反标签的数据增广方法提升性能,最终在3个比较难的自然语言处理任务上取得和full label相近的效果。
(3)包括PET、ptuing、ADAPET、EFL等目前主流的少样本算法以及在GPU上(CSL任务)的实验结果
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》