ãã£ã1人ããã¯ããããAgile Community of Practiceãï½ã½ã¼ã¹åçã¨Fearless Changeãæ·»ãã¦ï½
ããã«ã¡ã¯ãã»ãããã§ãï¼ å¼·åå¦ç¿ã£ã¦ç¥ã£ã¦ã¾ããï¼ ãAlphaGoããç¢ã®ä¸ççè ãç ´ã£ãã ãªã©ã¨æè¿èå ãæµ´ã³ã¦ããæ©æ¢°å¦ç¿æè¡ã§ããã ç§ã®ããã°ã§ãä½åãé¢é£è¨äºãåºãã¦ããã®ã§ããã ä»åã¯ãChainerã§å¼·åå¦ç¿ãç°¡åã«æ±ãããã¼ã«ããChainerRLãã使ã£ã¦ã¿ã¾ããï¼ ãªããªã便å©ã ã£ãã®ã§ä½¿ãæ¹ã解説ãã¾ã¨ããTipsãå ããªããã¡ã¢ãã¦ã¿ã¾ããã (ã³ã¼ãã¯jupyteræºæ ãªã®ã§ãä¸ããé çªã«ã³ããããã¨åºæ¬ã¯åãã¯ãã§ã) ããããå¼·åå¦ç¿ãã£ã¦ã¿ããã¨ãã人ã¯ãããã§å¼·åå¦ç¿ãã©ããªãã®ã試ãã¦ã¿ãã®ãããããããã¾ããï¼ å¼·åå¦ç¿ã£ã¦ï¼ chainerrl Setup å¿ è¦ãªã©ã¤ãã©ãªãimportãã environmentã®è¨å® environmentã«å¿ è¦ãªè¦ä»¶(æä½é) ä»å使ããã® Agentã®è¨å® ChainerRLã§ããã©ã§å®è£ ããã¦ãã
ï¾ï¾ï½°ï¾ï¾ï¾ï½°ï¾ï¾ï½¸ï¾ã¯ã³ã¢ãã£ãã£åãã¦ãã¦ãã³ãä»ãã®æ¹ãä»å 価å¤é«ãã¨ããããï¾ï¾-FNã§ä¸»ã«å·¥ä½æ å½ã®tai2anã§ãã NHKã§å ¨å½æ¾éãããAmazon Picking Challengeã§ã¬ã ãã¹ã£ãã¹ãã®ãã³ããããã³ã³ææºè¼ã®æ»ãå°ã¨ããå·¥ä½ãã¦ã¾ããã ã¨ã¯ããããã£ã±ãã¡ãã£ã¨ãã£ã¼ãã©ã¼ãã³ã°ãã¦ã¿ããã®ã§1,2ãæåããchainerãåå¼·ãå§ãã¾ããã ãã£ãããªã®ã§ç·ç»ã®çè²ãããããªã¼ã¨æã£ã¦è²ã 試ãã¦ã¿ã¾ããã ç·ç»ã®çè²ã¯æ師ããå¦ç¿ãªã®ã§ç·ç»ã¨çè²æ¸ã¿ã®ç»åã®ãã¼ã¿ã»ããã(ã§ããã°å¤§éã«)å¿ è¦ã§ãã ä»åã¯OpenCVã§ã«ã©ã¼ã®ç»åããç·ç»ãé©å½ã«æ½åºãã¦ãã¾ãã æ½åºä¾ â ã«ã©ã¼ã®ç»åãéãã¦ç·ç»ãä½ãã°ãã¼ã¿ã»ããã®å®æã§ããï¼ä»åã¯ï¼ï¼ä¸æããã使ã£ã¦ãã¾ãï¼ ãããã¯ã¼ã¯ã®å½¢ã§ãããU-netã¨ããæåã®æ¹ã§ã³ã³ããªã¥ã¼ã·ã§ã³ããæã®å±¤ã®åº
ãã®è¨äºã¯Chainer Advent Calendar 2016ã®14æ¥ç®ã®è¨äºã«ãªãã¾ãã æ¦è¦ ãã¡ãã®ãã¤ã¼ãã®ç 究ã«ã¤ãã¦ã詳ããã話ãã¾ãã æ°æµ·èª ããã®ä¸ç観ã大好ããªã®ã§ãããæ®å½±ããåçãæ°æµ·èª ãã風ã®ã¤ã©ã¹ãã«èªåå¤æã§ãããé¢ç½ããªã¨æããã·ã¹ãã ã試ä½ãã¦ã¿ã¾ããã ä¸ã®åçãå¤æããã®ãä¸ã®ã¤ã©ã¹ãã§ããã¯ãªãªãã£ã¯ã¾ã ã¾ã ã§ãããä¸æã使ãã°ä½¿ãããï¼èªåå¤æå¾ãè¥å¹²ã®æåè£æ£ãè¡ã£ã¦ãã¾ãï¼ pic.twitter.com/qY07hmzvHFâ null (@NrNrNr7) 2016å¹´9æ22æ¥ @NrNrNr7 æ¯è¼ç¨ã«æ ç»ã®ã·ã¼ã³ãå ¨ä½ã®è²åãã¯ããªãä¼¼ãããã¨ãã§ãã¦ããã®ã§ããã使ã£ã¦ããã¢ã«ã´ãªãºã ã®å¼±ç¹ã§é¨åçã«ããããªè²ã«ãªã£ã¦ãã¾ãã®ã課é¡ã§ãã ï¼ç»åã¯ããããè¨ã®èã®åºãç§é5ã»ã³ãã¡ã¼ãã«ãåã®åã¯ãããå¼ç¨ï¼ pic.twitter
Deleted articles cannot be recovered. Draft of this article would be also deleted. Are you sure you want to delete this article? Chainerã¯ãPreferred Networksãéçºãããã¥ã¼ã©ã«ãããã¯ã¼ã¯ãå®è£ ããããã®ã©ã¤ãã©ãªã§ãããã®ç¹å¾´ã¨ãã¦ã¯ã以ä¸ã®ãããªç¹ãããã¾ã(ãã¼ã ãã¼ã¸ãã)ã é«é: CUDAããµãã¼ãããGPUãå©ç¨ããé«éãªè¨ç®ãå¯è½ æè»: æè»ãªè¨æ³ã«ãããç³ã¿è¾¼ã¿ããªã«ã¬ã³ããªã©ãæ§ã ãªã¿ã¤ãã®ãã¥ã¼ã©ã«ããããå®è£ å¯è½ ç´è¦³ç: ãããã¯ã¼ã¯æ§æãç´è¦³çã«è¨è¿°ã§ãã å人çã«ã¯ãããã«ä¸ã¤ãã¤ã³ã¹ãã¼ã«ãç°¡åãã¨ããã®ãæãããã¨æãã¾ãã ãã£ã¼ãã©ã¼ãã³ã°ç³»ã®ãã¬ã¼ã ã¯ã¼ã¯ã¯ã©ããã¤ã³ã¹ãã¼ã«ãé¢åãªãã®ãå¤ãã®
æ¦è¦ Chainerã使ã£ã¦ç»åã®ãã£ãã·ã§ã³çæãå®è£ ãã¾ãããç»åãå ¥åããã¨ãã®èª¬ææãçæãã¾ããã½ã¼ã¹ã³ã¼ãã¯ä»¥ä¸ã«ããã¾ãã https://github.com/dsanno/chainer-image-caption 以ä¸ã®è«æã®ã¢ã«ã´ãªãºã ã使ãã¾ããã Show and tell: A neural image caption generator ãã§ã«Chainerã§ãã£ãã·ã§ã³çæãå®è£ ããã¦ããæ¹ãããã®ã§ããã¡ããåèã«ãã¾ããã Image caption generation by CNN and LSTM ï½ Satoshi's Blog from Bloomington ãã£ãã·ã§ã³çæã¢ãã« è«æã§ä½¿ç¨ãããã£ãã·ã§ã³çæã¢ãã«ã¯å¤§ããåãã¦3ã¤ã®ãããã¯ã¼ã¯ã§æ§æããã¦ãã¾ãã ç»åããã¯ãã«ã«å¤æãã${\rm CNN}$ ${\rm CNN}$ã«
ãã£ã¼ãã©ã¼ãã³ã°ã¯è²ã ãªç¥èãå¿ è¦ã§ãã ããã¥ã¼ããªã¢ã«åããã¾ãããã§æ¢ã¾ã£ãã¾ã¾ã®äººãå¤ãã§ãããã ããã§ãèªåãchainerãç解ããªããåãããæã®æ¹æ³ãæ¸ãã¾ãã 大åæ ã¨ã¯è¨ã£ããã®ã®ã誰ã§ãå ¥éã¯ç¡çãªã®ã§ã 以ä¸ãæä½éã®åæã§ãã æ°å¦:å¾®åç©åãè¡åã確çã¨ãèãã¦ããããããããã»ã»ã»ããããã«ã¯æã IT:Pythonããããªãã«åãããnumpyã¯ããããåããã æ©æ¢°å¦ç¿:説æå¤æ°ã¨ããç®çå¤æ°ã¨ãåé¡ã¨ããä½ã¨ãªãåãã ããæ°:å¼ã£ããã£ãæã«ã調ã¹ãæ°åããã ã´ã¼ã« chainerãèªåã§è²ã 試ããã¬ãã«ã«ãªãã ããã¾ã§è©¦ããã¬ãã«ã«ãªãã ãã§ãã ããã§ãåéã®éãä¸æ©ããã§ãã ä¸ç·ã«é å¼µãã¾ãããã ã¡ãªã¿ã«ãªãchainerããã¨ããã¨ã å ¥éè ã触ãã«ã¯ããã©ã³ã¹ã¯è¯ãããã«è¦ããããã§ãã æåã®æå ãã£ã¼ãã©ã¼ãã³ã°ã¯æ´å²ãã
PFIã»ããã¼(2016/02/25)ã§çºè¡¨ããã¹ã©ã¤ãã§ããä¼æ¿ãµã³ããªã³ã°å¯è½ãªçæã¢ãã«ã«é¢ããDeep Learningã®ç´¹ä»ã§ãï¼ãã¼ã¯ã¼ãï¼Wake-Sleep, å¤å AutoEncoder, Generative Adversarial Nets, Likelihood Ratioï¼Read less
æ¦è¦ ç³ã¿è¾¼ã¿ãã¥ã¼ã©ã«ãããã¯ã¼ã¯ã«ããç»åçæã¢ãã«ï¼DCGANï¼ã«å¼ç¤¾ã®ã¯ã³ãã¼ã¹ç»å10ä¸æãå¦ç¿ããã¾ããã å¾ãããçæã¢ãã«ã使ã£ã¦ãä¹±æ°ã§ä½ã£ã100次å ãã¯ãã«ããã¯ã³ãã¼ã¹ç»åãçæãã¾ããã éã«ãä¸æã®ã¯ã³ãã¼ã¹ç»åã100次å ãã¯ãã«ã«å§ç¸®ããå¯è¦åãã¾ããã å¯è¦åãããã¨ã§ãã¢ãã«ãã¯ã³ãã¼ã¹ã®ãè²ããå½¢ããæããã¢ãã«ããããã³ã®æç¡ããªã©ãåºæ¬çãªç¹å¾´ãæãããã¦ãããã¨ããããã¾ãã ãã®æè¡ã¯ãèªåã¿ã°ä»ããé¡ä¼¼ç»åæ¤ç´¢ã«å¿ç¨ãããã¨ãã§ãã¾ãã ã¯ããã« ã¯ããã¾ãã¦ããã¼ã¿ãµã¤ã¨ã³ã¹ãã¼ã ã®å¾è¤ã¨ç³ãã¾ããç¾å¨ãã¢ã¤ãã ç»åã®ã«ãã´ãªå¤å®ã¢ãã«ãä½ã£ãããåé¨éã®KPIã®æ¥ã ã®å¤åãã·ã¹ãã ã®ç°å¸¸ãç¥ãããããã·ã¥ãã¼ããä½ãä»äºããã¦ãã¾ãã ã¾ããææ°ã®ç 究è«æã«ãã¢ã³ãããå¼µããææ¡ææ³ã®æ¤è¨¼ã«ãåãçµãã§ãã¾ããä»åã¯ãå»å¹´ã®æ®ãããããã
2016/2/4 ãã½ããã¦ã§ã¢ã¸ã£ãã³ãããã°ãã¼ã¿æ´»ç¨å®åãã©ã¼ã©ã ãã§ã®ãã¬ã¼ã³è³æã§ãã 主ã«Pythonã§æ¸ããã³ã¼ãã®é«éåã®è©±ã§ããã¿ã¤ãã«ã¨ä¸èº«ããããããªãæããããã®ã¯èªè¦ãã¦ãã¾ããRead less
ããã¤ãåå¼·ããããã¨äººå·¥ç¥è½ï¼æ©æ¢°å¦ç¿ï¼ãã£ã¼ãã©ã¼ãã³ã°ï¼Deep Learningï¼ã¨ãã£ããããã¯ã®è¨äºã®è¦ã¤ãã¦ã¯ã¢ã¼ã«ã¤ããã¦ãããã®ã®ãçµå±2015å¹´ã¯ä½ä¸ã¤ãããã«çµãã£ã¦ãã¾ã£ãã®ã§ãã¨ã«ããä¸æ©ã§ã足ãè¸ã¿åºãã¹ãã æ¬è³ªçãªç解çã¯ãã¦ãããã¨ã«ãã試ãã¦ã¿ã ã¨ãããã¨ããã£ã¦ã¿ã¾ããã 試ããã®ã¯ãTensorFlowãChainerãCaffe ã¨ãã£ãæ©æ¢°å¦ç¿ããã³ãã£ã¼ãã©ã¼ãã³ã°ã®ä»£è¡¨çãªã©ã¤ãã©ãªï¼ãã¬ã¼ã ã¯ã¼ã¯3種ã¨ã2015å¹´ã«è©±é¡ã«ãªã£ããã£ã¼ãã©ã¼ãã³ã°ãå©ç¨ããã¢ããªã±ã¼ã·ã§ã³2種ï¼DeepDreamãchainer-goghï¼ã ï¼DeepDreamã§è©¦ããçµæç»åï¼ ã¿ã¤ãã«ã«åæ¥ã¨æ¸ãã¾ãããããã¨ãã°TensorFlowã¯ç°å¢æ§ç¯ã ããªã10åãããã°çµããã§ãããããChainerãªãã¦ã³ãã³ãä¸çºãªã®ã§5ç§ãããã§ããCaffeã¯å
ããã¤ãåå¼·ããããã¨äººå·¥ç¥è½ï¼æ©æ¢°å¦ç¿ï¼ãã£ã¼ãã©ã¼ãã³ã°ï¼Deep Learningï¼ã¨ãã£ããããã¯ã®è¨äºã®è¦ã¤ãã¦ã¯ã¢ã¼ã«ã¤ããã¦ãããã®ã®ãçµå±2015å¹´ã¯ä½ä¸ã¤ãããã«çµãã£ã¦ãã¾ã£ãã®ã§ãã¨ã«ããä¸æ©ã§ã足ãè¸ã¿åºãã¹ããæ¬è³ªçãªç解çã¯ãã¦ãããã¨ã«ãã試ãã¦ã¿ãã¨ãããã¨ããã£ã¦ã¿ã¾ããã 試ããã®ã¯ãTensorFlowãChainerãCaffe ã¨ãã£ãæ©æ¢°å¦ç¿ããã³ãã£ã¼ãã©ã¼ãã³ã°ã®ä»£è¡¨çãªã©ã¤ãã©ãªï¼ãã¬ã¼ã ã¯ã¼ã¯3種ã¨ã2015å¹´ã«è©±é¡ã«ãªã£ããã£ã¼ãã©ã¼ãã³ã°ãå©ç¨ããã¢ããªã±ã¼ã·ã§ã³2種ï¼DeepDreamãchainer-goghï¼ã ï¼DeepDreamã§è©¦ããçµæç»åï¼ ã¿ã¤ãã«ã«åæ¥ã¨æ¸ãã¾ãããããã¨ãã°TensorFlowã¯ç°å¢æ§ç¯ã ããªã10åãããã°çµããã§ãããããChainerãªãã¦ã³ãã³ãä¸çºãªã®ã§5ç§ãããã§ããCaffeã¯åã¯ã
Deleted articles cannot be recovered. Draft of this article would be also deleted. Are you sure you want to delete this article? PFNã®mattyaã§ããchainerã使ã£ãã¤ã©ã¹ãèªåçæããã£ã¦ã¿ã¾ãã(ä¸ã®ç»åããã®ä¸ä¾ã§ã)ã 20æ¥ç®ã®@rezoolabããã®è¨äº(Chainerã使ã£ã¦ã³ã³ãã¥ã¼ã¿ã«ã¤ã©ã¹ããæããã)ã¨ãã¿ã被ã£ã¡ãã£ãã®ã§ãæ¬è¨äºã§ã¯ããã«çºå±çãªã¨ãããæ¸ãã¦ããããã¨æãã¾ããä¸ç·ã«èªãã§ããã ãã¨ãããã¨ã æ¦è¦ Chainerã§ç»åãçæãããã¥ã¼ã©ã«ãããã§ããDCGANãå®è£ ããâgithub safebooruããé¡ã¤ã©ã¹ããéãã¦ãã¦å¦ç¿ããã å¦ç¿æ¸ã¿ã¢ãã«ãconvnetjsã§èªã¿è¾¼ã¾ãã¦ããã©ã¦ã¶ä¸ã§åãã
Deleted articles cannot be recovered. Draft of this article would be also deleted. Are you sure you want to delete this article? æ¦è¦ ã¤ãä¸ã¶æåã«ææ¡ããã深層å¦ç¿ã¢ãã«ã§ããDeep Convolutional Generative Adversarial Networks (以ä¸DCGAN)ãchainerä¸ã§å®è£ ããï¼ 70ä¸æãã®å¤§éã®ã¤ã©ã¹ãã使ã£ã¦DCGANã«ã¤ã©ã¹ãããããå¦ç¿ãããï¼ å¾ãããã¢ãã«ãå©ç¨ãã¦ï¼ã³ã³ãã¥ã¼ã¿ã«ã¤ã©ã¹ããæããã¦ã¿ãï¼ çµæã¨ãã¦ã¯çµæ§ä¸æãè¡ã£ãï¼æ¬ç©ã¨è¦åããã¤ããªãã¨ããã¬ãã«ã§ã¯ãªããã®ã®ï¼DCGANã¯æ£å¸¸ã«ã¤ã©ã¹ããããç»åãçæã§ãã¦ããï¼ ãã¿ãç大ã«è¢«ã£ã¦æ¬å½ã«ã¤ãã ã¯ããã« DCGANã¨ããç»å
ãã®è¨äºã¯ Chainer Advent Calendar 2015 17 æ¥ç®ã®è¨äºã§ãã ã¯ããã« ãµã¤ãºã大ãããã¼ã¿ã Deep Learning ããã¨å¦ç¿ã«æéãããã£ã¦ã¤ãããæéãããã£ã¦ã¤ããã®ã§ä¸¦åå¦çãã¦é«éåãããã 並ååããã®ã«è¯ããããªããã±ã¼ã¸ãªãããª? ã¨æ¢ãã¦ã¿ãã¨ãDask ã¨ãã並å / Out-Of-Core è¨ç®ããã±ã¼ã¸ãè¦ã¤ããããã㨠Chainer ãçµã¿åãããã¨ä¸¦åå¦çãç°¡åã«æ¸ããããªæ°ãããã æå㯠MNIST ã並ååãã¦ã¿ããããã¼ã¿ãå°ããããããã ãããé ããªã£ã¦ãã¾ã£ããããå°ã大ãããã¼ã¿ã§ãã CIFAR-10 ã使ããããæ·±ããããã¯ã¼ã¯æ§é ã§ãã®å¹æã確ããããã æçµçã«ã¯ä»¥ä¸äºã¤ã®å¦çã並ååãããã¨ãç®æãã Data Augmentation DNN ã®å¦ç¿ 1. Data Augmentation
çããããã«ã¡ã¯ ãå æ°ã§ãããç§ã¯å ¨ç¶ã§ãã Deep Learningãä¸è¨ã§è¨ãã¨ãã ã®æ·±å±¤å¦ç¿ã§ããã ä½ãæãç¨éã«ãã£ã¦æ§é ãå ¨ç¶éãã¾ãã ä»åã¯éå¼ãè¾å ¸ãããããDeep Learningã®å®è£ ã®ãªã³ã¯éãä½ã£ã¦ã¿ã¾ããã ä»åã¯ã©ã¤ãã©ãªã¯åãããæ²è¼ãã¾ãã CaffeãTheanoï¼Lasagneï¼ãTorch7ãChainerãªãã§ããããã§ãã å¾æ¥ã追è¨ããããã»ã»ã» Neural Networkï¼Full Connectedï¼ Auto Encoder Auto Encoder Denoising AutoEncoder Convolutional AutoEncoder Convolutional Neural Network Convolutional Neural Network R-CNN Fast-RCNN Faster-RCNN Recurren
2015 - 12 - 13 Chainerãç¨ãããã«ã¦ã§ã¢æ¤åºã¸ã®åãçµã¿ã«ã¤ã㦠ãã®è¨äºã¯ Chainer Advent Calendar 2015 13æ¥ç® ã®è¨äºã§ãï¼ ã¯ããã« Chainerã§å®è£ ãé²ãã¦ããã®ã§ããï¼ä»åã¯æ§æ³çãªè©±ããã¾ãï¼ã容赦ä¸ããï¼ ç§ã¯ 修士è«æ ã§Eãã£ã¹ã«ã㪠*1 ã対象㫠èªç¶è¨èªå¦ç 㨠æ©æ¢°å¦ç¿ ãç¨ãã¦ç 究ãé²ãã¦ãã¾ããï¼ããã¨ã¯å¥ã«ï¼Deep Learning㧠ãã«ã¦ã§ã¢ æ¤åºã«åãçµãã§ãã¾ãï¼æ¬å½ã¯ç 究室é å±ãããæã«ï¼ããã§è«ææ¸ãããã¨æã£ã¦ãã¾ãããï¼è²ã 大å¤ã§ãããã¨ãçºè¦ããã®ã§ï¼å人çã«ç´°ã ã¨ãã£ã¦ããæãã§ãï¼åæ¥ã¾ã§ã«ãªãã¨ãå®ç¾ãããã¨å¥®éãã¦ããã¾ãï¼ ãã«ã¦ã§ã¢ ã æ©æ¢°å¦ç¿ ãããããã®æ¦ç¥ ãã«ã¦ã§ã¢ ã«ãæ§ã ãªç¨®é¡ãããã¾ããï¼PEãã©ã¼ããã *2 ã® ãã«ã¦ã§ã¢ ã対象ã«ãã¦ããã¾ãï¼ ãã«
èªç¶è¨èªå¦çã¨ãã¥ã¼ã©ã«ããã ããæ°å¹´ã§ãèªç¶è¨èªå¦çã®åéã§ããã¥ã¼ã©ã«ããããé常ã«é »ç¹ã«ä½¿ãããããã«ãªã£ã¦ãã¾ããã èªç¶è¨èªå¦çã§ä¸»ã«è§£æ対象ã¨ãªãã®ã¯åèªã®é åãæ§ææ¨ãªã©ã§ããããã®å å ããæ å ±ã表ç¾ããããã«recurrent neural network1ãrecursive neural network1ãªã©ã«åºã¥ãã¢ãã«ãé »ç¹ã«ä½¿ããã¾ãããããã®æ大ã®ç¹å¾´ã¯ãã¥ã¼ã©ã«ãããããã種ã®ãã¼ã¿æ§é ãæã£ã¦ããã¨ããç¹ã§ã1ã¬ã¤ã¤ãããã®ãã¼ãã¯ããã»ã©å¤ããªã代ããã«ãããã¯ã¼ã¯ã®æ¥ç¶ãè¤éã§ããããå ¥åããããã¼ã¿ãã¨ã«ãããã¯ã¼ã¯ãã®ãã®ã®å½¢ç¶ãå¤åããã¨ããç¹å¾´ãããã¾ãããã®ãããä¼çµ±çãªfeedforward neural networkãåæã¨ãããã¼ã«ãããã§ã¯æ§ç¯ãé£ããã¨ããåé¡ãããã¾ããã Chainerã¯ããã®ãããªåé¡ãæ¦ã解決ãã¦ãã¾ãå¼·å
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}}