Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
slides
README.md

README.md

无监督学习论文列表

参考来源

目录

生成内容

A Neural Algorithm of Artistic Style

  • 传说中的Neural Style

Image Completion with Deep Learning in TensorFlow代码

  • 用GAN做图像修复(image inpainting任务),主要思想是同时优化两个目标:
  • 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)
  • 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)
  • 论文:Semantic Image Inpainting with Perceptual and Contextual Losses

生成对抗网络Generative Adversarial Network

Generative Adversarial Networks代码

  • Goodfellow的GAN开山之作

Conditional Generative Adversarial Nets

Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks代码

  • 生成房间图片

  • 戴眼镜男人-不戴眼镜男人+不戴眼镜女人=戴眼镜女人

  • 从一张人脸渐变到另一张人脸

  • “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”

Improved Techniques for Training GANs代码

  • 改变架构,解决GAN训练不稳定的问题
  • 半监督学习,少量标注样本,效果比Ladder Network还好一些

InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets代码

  • 对representation code空间施加一些要求,使其更具结构化,而非混沌一团
  • 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

变分自编码机Variation Auto Encoder

Pixel RNN类模型

Pixel Recurrent Neural Networks

Conditional Image Generation with PixelCNN Decoders

自编码机Auto Encoder

Stacked What-Where Auto-encoders

梯子网络Ladder Network

From neural PCA to deep unsupervised learning

  • 提出Ladder架构,但还未做半监督学习

Semi-Supervised Learning with Ladder Network

  • 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%

Deconstructing the Ladder Network Architecture

  • 深入挖掘Ladder Network的原理

One-shot Learning

One-Shot Generalization in Deep Generative Models

Zero-shot Learning

Biologically Plausible Learning

Towards Biologically Plausible Deep Learning

Towards a Biologically Plausible Backprop

Feedforward Initialization for Fast Inference of Deep Generative Networks is Biologically Plausible

其他

Towards Principled Unsupervised Learning

  • 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
You can’t perform that action at this time.