2. æ¸èªæ å ± ã¿ã¤ãã«ï¼Glow: Generative Flow with Invertible 1x1 Convolutions èè ï¼Diederik P. Kingma, Prafulla Dhariwal OpenAI, San Francisco -> æ稿æ¥ï¼2018/7/9 é¸å®çç±ï¼VAEã»Adamã®ææ¡è ã§ããD.P.Kingmaã®è«æ Flow-basedã®çæã¢ãã«ã¯èªãã ãã¨ããªãã£ã ⢠ä¸å¿NICE,realNVPãèªãã ã®ã§ç´¹ä»ãã¾ã ç»åçæãããã¨ãåæã£ã¦å¼ã³ã¾ã â»æars-pytorchã§å®è£ ããå§ãã¦ãã®ãè¦ã¦ããæ°ç¡ããã â»ééã£ã¦ããææãããããé¡ããã¾ã 2
ããã°ãã¯ï¼ ä»æ¥ã¯çµ±è¨ãæ©æ¢°å¦ç¿ã«ããã¦æãåºæ¬ã¨ãªãææ³ã§ããç·å½¢å帰ããåºçºãï¼ï¼ã¤ï¼ã¤ã¢ãã«ãå¦ç¿æ¹æ³ã«å¤æ´ãå ãã¦ããï¼æçµçã«æ·±å±¤å¦ç¿ã®åéã§é常ã«è¯ã使ããã¦ããçæã¢ãã«ã§ããå¤åãªã¼ãã¨ã³ã³ã¼ãï¼variational auto-encoderï¼VAEï¼*1*2ãå°ãã¦ããããã¨æãã¾ãï¼ 2014å¹´ã«çºè¡¨ãããVAEã¯ï¼å¾é è¿ä¼¼ãå¾ãããã®reparametrization trickãï¼å¹ççã«æ½å¨å¤æ°ãè¿ä¼¼æ¨è«ããèªèã¢ãã«ï¼recognition model, inference modelï¼ã®å°å ¥ï¼ç¢ºççå¾é æ³ï¼stochastic gradient descentï¼SGDï¼ã®2éé©ç¨ãªã©ï¼æ§ã ãªã¢ã¤ãã¢ãæ£ãã°ãããã¦ãã確çççæã¢ãã«ã§ãï¼èæ¯ã¨ãã¦ã¯ï¼å½æãã¥ã¼ã©ã«ãããã¯ã¼ã¯ãç¨ãã¦ç»åãçæããã¨ãã£ãã¢ããã¼ããï¼CNNã使ã£ãèå¥ãªã©ã¨æ¯ã¹ã¦ï¼ã
ä»åã¯ãVariational Autoencoder (VAE) ã®å®é¨ããã¦ã¿ããã å®ã¯èªåãå§ãã¦Deep Learningã«èå³ãæã£ãã®ããã®VAEãªã®ã ï¼VAEã®æ½å¨ç©ºéãããã£ã¦å¤æ§ãªé¡ç»åãçæãããã¢ï¼Morphing Facesï¼ãè¦ã¦ããããé³å£°åæã®å£°è³ªçæã«ä½¿ãããã¨æã£ãã®ãèå³ã®ãã£ããã ã£ãã ä»åã®å®é¨ã¯ãPyTorchã®å ¬å¼ã«ããVAEã®ã¹ã¯ãªãã ãèªåãªãã«èªã¿è§£ãã¦ã¾ã¨ãã¦ã¿ãçµæã«ãªã£ã¦ããã 180221-variational-autoencoder.ipynb - Google ãã©ã¤ã ãã£ããå®é¨ï¼ãã¤ãã®importã import os import numpy as np import torch import torch.nn as nn import torch.utils.data import torch.optim a
æ°ãããã¥ã¼ã©ã«ãããã¯ã¼ã¯ã®ã¢ã¼ããã¯ãã£ããã®æã ã§èªçããããããããå ¨é¨ãææ¡ãããã¨ã¯å°é£ã§ããå ¨ã¦ã®ç¥èªãè¦ãããã¨ããã¨ãæåã¯ãã®æ°ã®å¤ãã«å§åããã¦ãã¾ãã§ãããï¼DCIGNãBiLSTMãDCGANãç¥ã£ã¦ãã人ã¯ãã¾ããï¼ï¼ã ãããªããã§ããããã®ã¢ã¼ããã¯ãã£ã®å¤ããçãè¾¼ãã ãã¼ãã·ã¼ããä½ããã¨ã«ãã¾ããããã®ã»ã¨ãã©ã¯ãã¥ã¼ã©ã«ãããã¯ã¼ã¯ã§ããããããä¸ã«ã¯å ¨ãç°ãªãã¢ã¼ããã¯ãã£ãæ½ãã§ãã¾ããã©ããç¬ç¹ã§ç®æ°ããã¢ã¼ããã¯ãã£ã°ããã§ããããã¼ãã®æ§é ãæããã¨ã§åºæ¬çãªé¢ä¿ãåããããããªã£ã¦ãã¾ãã ãããããã¼ããããã¨ãã¦æããã¨ã®åé¡ç¹ã¯ãããããã©ã®ããã«ä½¿ãããããæ確ã«ç¤ºãã¦ããªãã¨ããç¹ã§ããä¾ãã°ãå¤åãªã¼ãã¨ã³ã³ã¼ãï¼VAEï¼ã¯ãªã¼ãã¨ã³ã³ã¼ãï¼AEï¼ã¨åãããã«è¦ãã¾ãããå®éã¯è¨ç·´éç¨ãå ¨ãç°ãªãã¾ãããè¨ç·´ãããããã¯ã¼ã¯ã®
Googleæ¥è±ç¿»è¨³ãNMT(ãã¥ã¼ã©ã«ãããã¯ã¼ã¯ãå©ç¨ãã翻訳ã¢ãã«)ã«ãªã£ããããã§ãã æèªãã¼ã¹ã§ããªããªç¿»è¨³ãã¦ããã¦ãã¿ããã§ã åèªã«ãã¦ã¹ãªã¼ãã¼å½ã¦ã¦ãåèªéã®å¯¾å¿é¢ä¿ãããããªããªã£ãã®ã¯ãã¿ãããã©ã ãããã«ç²¾åº¦ããããªã£ããããªããããªæ°ããããã ã¨ããããã§è±èªè«æãã£ã¡ãèªãã ããã VAEã§ä½ãããããªã¼ã¨æã£ã¦ã¦ãæç³»åãæ±ããVRAEå¨ãã調ã¹ã¾ããã Variational AutoEncoder Variational AutoEncoderãã¤ãã«ãã£ããç解ã§ããæ°ãããã ã¯ã£ã¤ã¼ããã®keraså®è£ ããã°ãã¨ã¦ãåå¼·ã«ãªãã¾ããã æé£ããããã¾ãã ralo23.hatenablog.com Variational AutoEncoderã«ã¤ãã¦ã¾ã¨ããã¹ã©ã¤ããä½ãã¾ããã Variational AutoEncoder from K
ã¯ããã« åºã¦ããå½åã¯ç»ååé¡ã¿ã¹ã¯ã§çå¨ãæ¯ãã£ã深層å¦ç¿ã§ãã, æè¿ã¯ããããªæ©æ¢°å¦ç¿ã¨çµã¿åããã§å¿ç¨ããã¦ãã¾ã. å¼·åå¦ç¿ãå¿ç¨ããAlphaGoã§ã¤ã»ã»ãã«ãæã¡è² ãããã, ç»åèªèã¨èªç¶è¨èªå¦çã®çµã¿åããã§ç»åã®ãã£ãã·ã§ã³ãçæããã, çæã¢ãã«ã«å¿ç¨ãã¦èªç¶ã«è¿ãç»åãä½ããªã©è³ãããè¦ãã¦ãã¾ã. ä»åã¯ç»åçæææ³ã®ãã¡ã®DeepLearningãèªç¶ã«çæã¢ãã«ã«æ¡å¼µããã¨èããããVAE(Variational Auto Encoder)ãã, ãã®çºå±ç³»ã§ããCVAE(Conditional VAE)ã¾ã§ã以ä¸ï¼ã¤ã®è«æããã¨ã«èªåã®æ¸ããkerasã®ã³ã¼ãã¨ã¨ãã«ç´¹ä»ãããã¨æãã¾ã. Auto-Encoding Variational Bayes Semi-Supervised Learning with Deep Generative Model
3. ICLR2016ã®ãã¬ã³ã Reinforcement Learning Unsupervised Learning Incorporating Structure Compressing Networks Incorporating Structure Initializing Networks Backprop Tricks Attention Deep Metric Learning Computer Vision Applications Visualizing Networks Do Deep Convolutional Nets Really Need to be Deep? Training-Free Methods Geometric Methods Gaussian Processes and Auto Encoders ResNet http://www.comp
IIBMP2016ï¼ç¬¬äºåçå½å»è¬æ å ±å¦é£å大ä¼ï¼ã§ã®æå¾ è¬æ¼ã®å 容ã§ãã深層å¦ç¿ãçæéç¨ã®åé¡ã¨ãã¦ã¨ããã¦ããªã表ç¾ãå¦ç¿ã§ããã®ãã説æããå¾ã«ï¼æ·±å±¤çæã¢ãã«ï¼VAE, GAN, å°ãã ãèªå·±å帰ã¢ãã«, ã¨ãã«ã®ã¼ã¢ãã«, ã¢ã¼ã¡ã³ããããã³ã°ã¢ãã«ãç´¹ä»ãã¾ããRead less
1. Ishikawa Watanabe Lab THE UNIVERSITY OF TOKYO http://www.k2.t.u-tokyo.ac.jp/ ç«ã§ãåãã Variational AutoEncoder 2016/07/30 é¾é ç¿ (Sho Tatsuno) 2. Ishikawa Watanabe Lab http://www.k2.t.u-tokyo.ac.jp/ ä»åã®å 容 ⢠Variational Auto-Encoderã®è§£èª¬ â çæã¢ãã«ãã®ãã®ã®æ¦è¦ â Variational Auto-Encoder(VAE)ã®ãªãã¹ãåã¿ç ãã解説 â ãã®ä»çæã¢ãã«è«æã®ãã£ããããç´¹ä» â¢ èª¬æãããã¨/ããªãã㨠â 説æããã㨠» çæã¢ãã«ã®ç°¡åãªæ¦è¦ã¨äºä¾ » Variational AutoEncoderã®æ§é ã¨æ°å¼çã»ç´æçç解 â 説æããª
ã©ã³ãã³ã°
ã©ã³ãã³ã°
ã©ã³ãã³ã°
ãªãªã¼ã¹ãé害æ å ±ãªã©ã®ãµã¼ãã¹ã®ãç¥ãã
ææ°ã®äººæ°ã¨ã³ããªã¼ã®é ä¿¡
å¦çãå®è¡ä¸ã§ã
j次ã®ããã¯ãã¼ã¯
kåã®ããã¯ãã¼ã¯
lãã¨ã§èªã
eã³ã¡ã³ãä¸è¦§ãéã
oãã¼ã¸ãéã
{{#tags}}- {{label}}
{{/tags}}