2016/03/20

生成モデル(Generative Model)関連の論文まとめ

2014年に出されたVAE論文やGAN論文などに加え、Neural StyleDeep Dream等のCNNによる画像生成も後押しして盛り上がっている生成モデル周辺ですが、関連する論文をまとめてみました。

ちなみに、生成モデル全般に関する概要はこの辺りの資料がまとまっていてオススメです。
[1] Semi-Supervised Learning with Ladder Network
[2] 生成モデルのDeep Learning
[3] 深層生成モデルによる表現学習
[4] Deep Learning - Chapter 20 : Deep Generative Models
[5] Open AI Blog - Generative Models
[6] Building Machines that Imagine and Reason
[7] Tutorial on Variational Autoencoder (VAEの分かりやすい解説)
[8] Comparison of different GAN Variants (様々なGANの拡張のモデル概要)
[9] NIPS 2016 Tutorial: Generative Adversarial Networks (GANのチュートリアル論文)

(最終更新 : 2017/2/8)

●Variational Autoencoder
・Auto-Encoding Variational Bayes, 2014, ICLR
・Neural Variational Inference and Learning In Belief Network, 2014, ICML
・Stochastic Backpropagation and Approximate Inference in Deep Generative Models, 2014, ICML
・Semi-Supervised Learning with Deep Generative Models, 2014, NIPS
・DRAW: A Recurrent Neural Network For Image Generation, 2015, ICML
・Deep Convolutional Inverse Graphics Networks, 2015, NIPS
・A Recurrent Latent Variable Model for Sequential Data, 2015, NIPS
・Learning Structured Output Representation using Deep Conditional Generative Models, 2015, NIPS
・Variational Dropout and the Local Reparameterization Trick, 2015, NIPS
・Generating Images from Captions with Attention, 2016, ICLR
・The Variational Fair Autoencoder, 2016, ICLR
・One-Shot Generalization in Deep Generative Models, 2016, ICML
・Autoencoding Beyond Pixels Using a Learned Similarity Metric, 2016, ICML
・Auxiliary Deep Generative Models, 2016, ICML
・Note On The Equivalence of Hierarchical Variational Models and Auxiliary Deep Generative Models, 2016, Arxiv
・An Uncertain Future: Forecasting from Static Images using Variational Autoencoders, 2016, ECCV
・Attend, Infer, Repeat: Fast Scene Understanding with Generative Models, 2016, NIPS
・Composing Graphical Models with Neural Networks for Structured Representations and Fast Inference, 2016, NIPS
・Ladder Variational Autoencoders, 2016, NIPS
・Towards Conceptual Compression, 2016, NIPS
・Unsupervised Learning of 3D Structure from Images, 2016, NIPS
・Variational Autoencoder for Deep Learning of Images, Labels and Captions, 2016, NIPS
・The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables, 2017, ICLR
・Categorical Reparameterization with Gumbel-Softmax, 2017, ICLR
・Reparameterization Trick for Discrete Variables, 2016, Arxiv
・Variational Lossy Autoencoder, 2017, ICLR
・PixelVAE : A Latent Variable Model for Natural Images, 2017, ICLR
・β-VAE : Learning Basic Visual Concepts With a Constrained Variational Framework, 2017, ICLR
・Stick-Breaking Variational Autoencoders, 2017, ICLR
・Discrete Variational Autoencoders, 2017, ICLR

●Improving Variational Lower Bound
・Markov Chain Monte Carlo and Variational Inference : Bridging The Gap, 2015, ICML
・Variational Inference with Normalizing Flows, 2015, ICML
・Importance Weighted Autoencoders, 2016, ICLR
・Variational Gaussin Process, 2016, ICLR
・Variationally Auto-Encoded Deep Gaussian Processes, 2016, ICLR
・Variational Inference for Monte Carlo Objectives, 2016, ICML
・Improving Variational Inference with Inverse Autoregressive Flow, 2016, NIPS
・Operator Variational Inference, 2016, NIPS
・Rényi Divergence Variational Inference, 2016, NIPS
・Variational Bayes on Monte Carlo Steroids, 2016, NIPS

●Generative Adversarial Networks
・Generative Adversarial Nets, 2014, NIPS
・Deep Generative Image Models Using a Laplacian Pyramid of Adversarial Networks, 2015, NIPS
・Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks, 2016, ICLR
・Unsupervised and Semi-Supervised Learning with Categorical Generative Adversarial Networks, 2016, ICLR
・Deep Multi-Scale Video Prediction Beyond Mean Square Error, 2016, ICLR
・Adversarial Autoencoders, 2016, ICLR Workshop
・Deep Directed Generative Models with Energy-Based Probability Estimation, 2016, ICLR Workshop
・Autoencoding Beyond Pixels Using a Learned Similarity Metric, 2016, ICML
・Generative Adversarial Text to Image Synthesis, 2016, ICML
・Pixel-Level Domain Transfer, 2016, ECCV
・Generative Visual Manipulation on the Natural Image Manifold, 2016, ECCV
・Precomputed Real-Time Texture Synthesis with Markovian Generative Adversarial Networks, 2016, ECCV
・Generative Image Modeling using Style and Structure Adversarial Networks, 2016, ECCV
・Improved Techniques for Training GANs, 2016, NIPS
・InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets, 2016, NIPS
・f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization, 2016, NIPS
・Generating Images with Perceptual Similarity Metrics based on Deep Networks, 2016, NIPS
・Coupled Generative Adversarial Networks, 2016, NIPS
・Synthesizing the Preferred Inputs for Neurons in Neural Networks via Deep Generator Networks, 2016, NIPS
・Learning What and Where to Draw, 2016, NIPS
・Generating Videos with Scene Dynamics, 2016, NIPS
・Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling, 2016, NIPS
・Generative Adversarial Imitation Learning, 2016, NIPS
・Connecting Generative Adversarial Networks and Actor-Critic Methods, 2016, NIPS Workshop
・Energy-based Generative Adversarial Network, 2017, ICLR
・Calibrating Energy-based Generative Adversarial Networks, 2017, ICLR
・Learning in Implicit Generative Models, 2017, ICLR Workshop
・Generative Adversarial Nets from a Density Ratio Estimation Perspective, 2016, Arxiv
・Amortised MAP Inference for Image Super-resolution, 2017, ICLR
・Towards Principled Methods for Training Generative Adversarial Networks, 2017, ICLR
・Improving Generative Adversarial Networks with Denoising Feature Matching, 2017, ICLR
・Mode Regularized Generative Adversarial Networks, 2017, ICLR
・LR-GAN: Layered Recursive Generative Adversarial Networks for Image Generation, 2017, ICLR
・Unrolled Generative Adversarial Networks, 2017, ICLR
・Generative Multi-Adversarial Networks, 2017, ICLR
・Adversarially Learned Inference, 2017, ICLR
・Adversarial Feature Learning, 2017, ICLR
・Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network, 2016, Arxiv
・Image-to-Image Translation with Conditional Adversarial Networks, 2016, Arxiv

●Adversarial Examples
・Intriguing Properties of Neural Networks, 2014, ICLR
・Explaining and Harnessing Adversarial Examples, 2015, ICLR
・Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images, 2015, CVPR
・Distributional Smoothing by Virtual Adversarial Examples, 2016, ICLR
・DeepFool: A Simple and Accurate Method To Fool Deep Neural Networks, 2016, CVPR
・Robustness of Classifiers: From Adversarial To Random Noise, 2016, NIPS
・Adversarial Training Methods for Semi-Supervised Text Classification, 2017, ICLR
・On Detecting Adversarial Perturbations, 2017, ICLR
・Delving into Transferable Adversarial Examples and Black-box Attacks, 2017, ICLR
・Adversarial Machine Learning at Scale, 2017, ICLR
・Adversarial Examples In The Physical World, 2017, ICLR

●生成モデルに関するその他の論文
・Training Deep Generative Models : Variations on a Theme, 2015, NIPS Workshop
・A Note On The Evaluation of Generative Models, 2016, ICLR
・A Test of Relative Similarity For Model Selection in Generative Models, 2016, ICLR
・Sampling Generative Networks, 2016, NIPS
・On the Quantitative Analysis of Decoder-Based Generative Models, 2017, ICLR

●The Helmholtz Machine
・Autoencoders, Minimum Description Length, and Helmholtz Free Energy, 1993, NIPS
・The Wake-Sleep Algorithm for Unsupervised Neural Networks, 1995, Science
・The Helmholtz Machine, 1995, Neural Computation
・Varieties of Helmholtz Machine, 1996, Neural Networks
 (その他Hintonが1995年前後に書いたHelmholtz Machineに関する論文はこの辺り)
・Reweighted Wake-Sleep, 2015, ICLR
・Bidirectional Helmholtz Machines, 2016, ICML

●その他の生成モデル
・The Neural Autoregressive Distribution Estimator, 2011, AISTATS
・Generalized Denoising Auto-Encoders As Generative Models, 2013, NIPS
・Deep Generative Stochastic Networks Trainable by Backprop, 2014, ICML
・Deep AutoRegressive Networks, 2014, ICML
・On the Equivalence Between Deep NADE and Generative Stochastic Networks, 2014, ECMLPKDD
・MADE: Masked Autoencoder for Distribution Estimation, 2015, ICML
・Generative Moment Matching Networks, 2015, ICML
・Pixel Recurrent Neural Networks, 2016, ICML
・Conditional Image Generation with PixelCNN Decoders, 2016, NIPS
・Wavenet: A Generative Model for Raw Audio, 2016, Arxiv
・PixelCNN++: Improving the PixelCNN with Discretized Logistic Mixture Likelihood and Other Modifications, 2017, ICLR

0 件のコメント:

コメントを投稿