Generation vs. Network Gan Rollup _ machine learning

Review and summary of the related articles on generative antagonism learning (generative adversarial network, GAN).

Article: Generative adversarial Nets (2014) [Paper][code]
Ian Goodfellow's first article about generative confrontation learning, groundbreaking work.
-This paper proposes to estimate the generation model by the confrontation network.
-The theory expounds the loss function of the model and its training method. The code has been integrated into the machine learning framework such as Theano.

Unsupervised representation Learning with Deep convolutional generative adversarial Networks (2015) [Paper][code] In this paper, a deep convolution generation countermeasure network (Dcgans) is proposed, which combines the Generative network (Gans) with the convolution Neural Network (CNNs). The trained discriminant model is used for image classification, and the results of other unsupervised methods are comparable. Photo-realistic single Image super-resolution using a generative adversarial
Network (2016) [paper] [code]
use Gan to finish To the high-definition image of low resolution. The successful application of Gan in the picture high-definition problem is clear, the effect is remarkable, very has the reference. Stackgan:text to Photo realistic Image synthesis with stacked generative adversarial network (2016) [Paper][code]
Gan for Image generation based on text description. Two gan,gan-i are used to generate 64x64 low-resolution images (predecessors have worked), gan-ii generate 256x256 high-resolution details with more images based on text and low-resolution images. It looks like a. gan text generation picture plus a picture of Gan High-definition, this is two independent research work, but the author explains the difference inside. Gan-ii also incorporates text descriptions, with additional information input, unlike images with no extra information. NIPS 2016 tutorial:generative adversarial Networks (2016) [paper]
Great God Goodfellow's summary after NIPS2016 Tutorial and the answers to important questions raised at the meeting. The summary is very strong, the content is many (57 pages), the dry goods are many. The content mainly has the following several aspects:
–why Study generative modeling?
–how do generative models work?
–how do gans work?
–tips and Tricks
–research Frontiers

Other:
1. Blog: gan Change world!–2016 A summary of the development of Gan in the year and some prospects for the 2017.