按年被提到
20162016 · 7 次2026
机器学习与深度学习›神经科学启发的 AI·arXiv:1605.06065 · cs.LG

One-shot Learning with Memory-Augmented Neural Networks

用记忆增强神经网络做单样本学习
Adam Santoro, Sergey Bartunov, Matthew Botvinick, Daan Wierstra, Timothy Lillicrap · 2016

用外部记忆让网络看几个样本就能学会新类别

开始阅读 ↗PDFalphaXiv 讨论
被推荐 · 8 人首次 2016-05-20 · 最近 2018-06-03
8不同的人
5活跃月份
8讨论串
384被引(OpenAlex)

单样本学习对深度网络仍是难题:传统基于梯度的网络需要大量数据和反复迭代,遇到新数据时还要低效地重学参数并易发生灾难性干扰。论文展示,带外部记忆的记忆增强神经网络(如神经图灵机 NTM)能快速吸收新数据,并在只见到少量样本后做出准确预测;作者还提出一种只基于内容、不再使用位置聚焦机制来访问外部记忆的新方法。

HN 为什么提它

2016 年的评论者用它反驳「神经网络需要海量数据」的观点,作为网络能从少量样本学习的新进展,也有人指出相关方法扩展性不佳,并提到 DeepMind 当时的最好结果。

方法难度 4 / 5约 35 分钟经得起时间
读之前最好知道
1神经图灵机(Neural Turing Machine)
2循环神经网络与 LSTM
3元学习(meta-learning)
4单样本与少样本学习

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
推荐dave_sullivan · 2016-06

“Here are 2 good papers--one from last year, one very recent--that might change your mind.”

推荐两篇近期好论文,用来反驳对进展的怀疑
The truth about deep learning原评论 ↗
引用jimfleming · 2016-06

“Newer work is even able to learn from just a few examples[0] and then generalize beyond that.”

指出新工作仅凭几个样本就能学习并泛化
AI, Deep Learning, and Machine Learning: A Primer [video]原评论 ↗
引用ak4g · 2016-11

“This is just an outdated view of the state-of-the-art. It's understandable, given that it's outdated by maybe six months, iff you're willing to go with preprints.”

反驳对方观点已过时,指出预印本里有较新进展
Show HN: DeepRhyme (D-Prime) – Generating dope rhymes with machine learning原评论 ↗
引用YeGoblynQueenne · 2016-11

“A paper someone pointed me to recently on HN suggested that it doesn't scale very well:”

提到有论文表明其扩展性不佳
Embed, encode, attend, predict: the new deep learning formula for NLP models原评论 ↗
引用vonnik · 2016-12

“And yet some great research is being done that will prove him wrong on his first point.”

称有出色研究会证明对方的第一点是错的
This AI Boom Will Also Bust原评论 ↗
引用nl · 2016-12

“I think the SOTA is probably [2], which came out of DeepMind.”

认为当前最优结果可能是 DeepMind 的这项工作
Open-sourcing DeepMind Lab原评论 ↗
看全部 7 条 →
引用goolulusaurs · 2018-06

“including: [LINK], One-shot Learning with Memory-Augmented Neural Networks and [OTHER_PAPER].”

列出受 Neural Turing Machine 启发的相关工作
Deep learning: a critical appraisal原评论 ↗
被提交为 story · 1 次