├── README.md └── slides ├── 第01讲-刘思聪-GAN下半部分.pdf ├── 第01讲-郑华滨-GAN上半部分.pdf ├── 第02讲-黄正杰-LinearFactorModel.pdf ├── 第03讲-刘思聪-Autoencoder.pdf └── 第04讲-郑华滨-RepresentationLearning.pdf /README.md: -------------------------------------------------------------------------------- 1 | # 无监督学习论文列表 2 | 3 | ## 参考来源 4 | - [Most Cited Deep Learning Papers # Unsupervised | Github](https://github.com/terryum/awesome-deep-learning-papers#unsupervised) 5 | - [从OpenAI看深度学习研究前沿 | 知乎专栏](https://zhuanlan.zhihu.com/p/20924929?f3fb8ead20=2fe7890562ecdbf5998ce5a6c0a1ba08) 6 | - [Generative Models | OpenAI博客](https://openai.com/blog/generative-models/) 7 | - [A Path to Unsupervised Learning through Adversarial Networks](https://code.facebook.com/posts/1587249151575490/a-path-to-unsupervised-learning-through-adversarial-networks/) 8 | - [Generative Adversarial Networks（GAN）的现有工作 | 程序媛的日常](http://chuansong.me/n/317902651864) 9 | 10 | ## 目录 11 | - [生成内容](#生成内容) 12 | - [生成对抗网络Generative Adversarial Network](#生成对抗网络generative-adversarial-network) 13 | - [变分自编码机Variation Auto Encoder](#变分自编码机variation-auto-encoder) 14 | - [Pixel RNN类模型](#pixel-rnn类模型) 15 | - [自编码机Auto Encoder](#自编码机auto-encoder) 16 | - [梯子网络Ladder Network](#梯子网络ladder-network) 17 | - [One-shot Learning](#one-shot-learning) 18 | - [Zero-shot Learning](#zero-shot-learning) 19 | - [Biologically Plausible Learning](#biologically-plausible-learning) 20 | - [其他](#其他) 21 | 22 | ### 生成内容 23 | [**A Neural Algorithm of Artistic Style**](https://arxiv.org/abs/1508.06576) 24 | - 传说中的Neural Style 25 | 26 | [**Image Completion with Deep Learning in TensorFlow**](http://bamos.github.io/2016/08/09/deep-completion/) 27 | （[代码](https://github.com/bamos/dcgan-completion.tensorflow)） 28 | - 用GAN做图像修复（image inpainting任务），主要思想是同时优化两个目标： 29 | - 1.原图中有完好区域和丢失区域，要让生成的修复图与原图在对应的完好区域尽可能接近（所谓Contextual Loss） 30 | - 2.要让生成的修复图尽可能被GAN的判别器判定为真实图片，尽可能像真的（所谓Perceptual Loss） 31 | - 论文：[**Semantic Image Inpainting with Perceptual and Contextual Losses**](https://arxiv.org/abs/1607.07539) 32 | 33 | ### 生成对抗网络Generative Adversarial Network 34 | 35 | [**Generative Adversarial Networks**](http://arxiv.org/abs/1406.2661) 36 | （[代码](https://github.com/goodfeli/adversarial)） 37 | - Goodfellow的GAN开山之作 38 | 39 | [**Conditional Generative Adversarial Nets**](https://arxiv.org/abs/1411.1784) 40 | 41 | [**Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks**](https://arxiv.org/abs/1511.06434) 42 | （[代码](https://github.com/Newmu/dcgan_code)） 43 | - 生成房间图片 44 | - 戴眼镜男人－不戴眼镜男人＋不戴眼镜女人＝戴眼镜女人 45 | - 从一张人脸渐变到另一张人脸 46 | 47 | - “这篇论文的提出看似并没有很大创新，但其实它的开源代码现在被使用和借鉴的频率最高……这些工程性的突破无疑是更多人选择 DCGAN 这一工作作为 base 的重要原因” 48 | 49 | [**Improved Techniques for Training GANs**](https://arxiv.org/abs/1606.03498) 50 | （[代码](https://github.com/openai/improved-gan)） 51 | - 改变架构，解决GAN训练不稳定的问题 52 | - 半监督学习，少量标注样本，效果比Ladder Network还好一些 53 | 54 | [**InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets**](https://arxiv.org/abs/1606.03657) 55 | （[代码](https://github.com/openai/InfoGAN)） 56 | - 对representation code空间施加一些要求，使其更具结构化，而非混沌一团 57 | - 结果在representation向量的单个维度上获得了非常好的可解释性，例如渐变一个维度的数值，生成的人脸图谱从“抬头姿态”到“低头姿态”渐变，非常像流形学习里面的一些例子 58 | 59 | ### 变分自编码机Variation Auto Encoder 60 | 61 | ### Pixel RNN类模型 62 | [**Pixel Recurrent Neural Networks**](http://arxiv.org/abs/1601.06759) 63 | 64 | [**Conditional Image Generation with PixelCNN Decoders**](http://arxiv.org/abs/1606.05328) 65 | 66 | ### 自编码机Auto Encoder 67 | 68 | [**Stacked What-Where Auto-encoders**](https://arxiv.org/abs/1506.02351) 69 | 70 | ### 梯子网络Ladder Network 71 | 72 | [**From neural PCA to deep unsupervised learning**](https://arxiv.org/abs/1411.7783) 73 | - 提出Ladder架构，但还未做半监督学习 74 | 75 | [**Semi-Supervised Learning with Ladder Network**](https://arxiv.org/abs/1507.02672) 76 | - 半监督学习，MNIST用100个标注数据达到约99%，CIFAR用4000个标注数据达到约80% 77 | 78 | [**Deconstructing the Ladder Network Architecture**](http://arxiv.org/abs/1511.06430) 79 | - 深入挖掘Ladder Network的原理 80 | 81 | ### One-shot Learning 82 | [**One-Shot Generalization in Deep Generative Models**](http://arxiv.org/abs/1603.05106) 83 | 84 | ### Zero-shot Learning 85 | 86 | ### Biologically Plausible Learning 87 | [**Towards Biologically Plausible Deep Learning**](http://arxiv.org/abs/1502.04156) 88 | 89 | [**Towards a Biologically Plausible Backprop**](http://arxiv.org/abs/1602.05179) 90 | 91 | [**Feedforward Initialization for Fast Inference of Deep Generative Networks is Biologically Plausible**](http://arxiv.org/abs/1606.01651) 92 | 93 | 94 | ### 其他 95 | 96 | [**Towards Principled Unsupervised Learning**](http://arxiv.org/abs/1511.06440) 97 | - 用GAN做半监督学习的论文中所定义的新的损失函数与这篇提出的Output Distribution Matching (ODM) cost有紧密联系 98 | -------------------------------------------------------------------------------- /slides/第01讲-刘思聪-GAN下半部分.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/WarBean/Unsupervised-Learning-Research/1cc64fcdac42a888814b203bc904cff182c506a5/slides/第01讲-刘思聪-GAN下半部分.pdf -------------------------------------------------------------------------------- /slides/第01讲-郑华滨-GAN上半部分.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/WarBean/Unsupervised-Learning-Research/1cc64fcdac42a888814b203bc904cff182c506a5/slides/第01讲-郑华滨-GAN上半部分.pdf -------------------------------------------------------------------------------- /slides/第02讲-黄正杰-LinearFactorModel.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/WarBean/Unsupervised-Learning-Research/1cc64fcdac42a888814b203bc904cff182c506a5/slides/第02讲-黄正杰-LinearFactorModel.pdf -------------------------------------------------------------------------------- /slides/第03讲-刘思聪-Autoencoder.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/WarBean/Unsupervised-Learning-Research/1cc64fcdac42a888814b203bc904cff182c506a5/slides/第03讲-刘思聪-Autoencoder.pdf -------------------------------------------------------------------------------- /slides/第04讲-郑华滨-RepresentationLearning.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/WarBean/Unsupervised-Learning-Research/1cc64fcdac42a888814b203bc904cff182c506a5/slides/第04讲-郑华滨-RepresentationLearning.pdf --------------------------------------------------------------------------------