按年被提到
20112012 · 3 次2026
视觉、语音与图像生成›卷积网络与图像识别·arXiv:1112.6209 · cs.LG

Building high-level features using large scale unsupervised learning

用大规模无监督学习构建高层特征
Quoc V. Le, Marc'Aurelio Ranzato, Rajat Monga, Matthieu Devin, Kai Chen, Greg S. Corrado, Jeff Dean, Andrew Y. Ng · 2011

Google 用 1.6 万核无监督训练,网络自发学出人脸和猫脸检测器

开始阅读 ↗PDFalphaXiv 讨论
被推荐 · 7 人首次 2012-05-30 · 最近 2023-03-19发表于 IEEE International Conference on Acoustics Speech and Signal Processing
7不同的人
6活跃月份
7讨论串
1,695被引(OpenAlex)

论文探讨能否仅用无标注数据学出类别特定的高层特征检测器:训练一个 9 层的局部连接稀疏自编码器(含池化和局部对比度归一化),拥有 10 亿连接,数据为 1000 万张网络图像,在 1000 台机器(1.6 万核)上用模型并行与异步 SGD 训练三天。结果表明无需标注也能得到人脸检测器,对平移、缩放和平面外旋转都有鲁棒性,还对猫脸和人体敏感;以这些特征为起点,在 20000 类的 ImageNet 识别上达到 15.8% 准确率,相对此前最优提升 70%。

HN 为什么提它

HN 上的人把它当作“Google 的猫脸实验”的出处:网络没有任何标签就学出了猫这个概念,甚至有人因此决定转行读博做 AI 研究。

实证研究难度 3 / 5约 40 分钟经得起时间
读之前最好知道
1自编码器(autoencoder)
2无监督学习
3卷积与池化
4随机梯度下降(SGD)
5分布式训练

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visarga · 2015-08

“Wasn't it amazing that they could distill the concept of cat from images with no help from external labels”

感叹它能不靠外部标签从图像中提炼出“猫”的概念
M, a personal digital assistant inside Facebook Messenger原评论 ↗
引用yk · 2012-06

“I did only print it (yet), but from the abstract it appears that they did mainly tried face recognition, not trying to find cats.”

只打印未读,据摘要看主要做的是人脸识别而非找猫
In a Big Network of Computers, Evidence of Machine Learning原评论 ↗
引用cs702 · 2012-06

“Paper here: Building high-level features using large scale unsupervised learning”

贴出论文:用无标注图像学出人脸特征,无需人工训练
Google & Stanford create a digital brain that learns to identify a human face原评论 ↗
引用varelse · 2013-08

“Given the 15.8% success rate of the cat detector on large data sets: Building high-level features using large scale unsupervised learning,”

引用猫检测器在大数据集上 15.8% 的成功率
The Face Scan Arrives原评论 ↗
引用nl · 2013-12

“Unsupervised image feature extraction[1]”

把它当作 Google 在无监督图像特征提取上的贡献
Google is the new Bell Labs原评论 ↗
引用p1esk · 2023-03

“Fast forward to 2012, when I saw this paper discussed in the news: Building high-level features using large scale unsupervised learning - that was when I decided to quit my job and enroll in a phd program to do AI research.”

称 2012 年在新闻里看到这篇论文,促使自己辞职读 PhD 做 AI 研究
Ask HN: I'm feeling awe, unsettled and disoriented by GPT-4. How about you?原评论 ↗
被提交为 story · 1 次