8æã®é ãããã£ã¼ãã©ã¼ãã³ã°ãå®è£
ãã¦ããã®ããå
æ¥ããã¬ã¼ã³ãã¦ãã¾ããã
プログラマのための数学勉強会@福岡 - connpass
â»è¿½è¨ 2023/4/12
SpeakerDeckã«ãç½®ãã¦ã¾ã
https://speakerdeck.com/kishida/weakest-deep-learning-i-implemented
GPU対å¿ããããããããã¢ã¦ãã¨ããããããã¨ãããããå®è£
ãã¦ãçµæ§ã¤ãããªã£ã¦ã¾ãã
ã¡ããã¨å¦ç¿ãã¦ãããªããã¨ä»¥å¤ã¯ã
ã½ã¼ã¹ã¯ãããªæãã«ãªã£ã¦ãã¦ãã¾ãã
https://github.com/kishida/neuralnet/tree/CorrectOperationAsCCN
GPU対å¿ã«ã¯aparapiã使ã£ã¦ãã¾ããJavaã§GPUã³ã¼ããæ¸ããã¹ã°ã¬ã¢ãã§ãã
aparapi - API for data parallel Java. Allows suitable code to be executed on GPU via OpenCL. - Google Project Hosting
ãã©ããã°ããã£ã¦ã¯ã¾ã£ã¦ãã®ã§ãä¿®æ£ãã¦ãããæãã¦ã¾ãã
Add parenthesis around conditional expression by kishida · Pull Request #10 · aparapi/aparapi
æè¿åæ»ãã¦ãã£ã½ãã¦ãåãè¾¼ã¾ããæ°ãããªãã
â» 9/9 åãè¾¼ã¾ãã¾ããï¼
åèæç®ã¨ãã¦ã¯ãã¨ãããããã®æ¬ããã£ã¼ãã©ã¼ãã³ã°ã«ã¤ãã¦ãã ããããããã¾ããåºæ¬çã«ã¯ããã®æ¬ãåèã«ãã¦å®è£ ãã¾ããã
ç´°ãããã©ã¡ã¼ã¿ããã¥ã¼ã©ã«ãããã¯ã¼ã¯ã®æ§æãªã©ã¯ããã¥ã¼ã©ã«ããããæµè¡ãã ãããã£ãã«ãªã£ãããã®è«æãåèã«ãã¦ã¾ããImageNet Classification with Deep Convolutional Neural Networks
ãã®ãµãã¤ãèªãã¨ãå®è£ ã«å¿ è¦ãªæ å ±ã¯ã ãããæãã¾ãã
å¦ç¿ç¨ã®ç»åã«ã¯ãCALTECH256ã¨ããã®ã使ã£ã¦ãã¾ãã
http://www.vision.caltech.edu/Image_Datasets/Caltech256/
â» 9/9è¿½è¨ ã«ãã´ãªãã¨ã®ç»åæ°ãã80æã ã£ãã700æã ã£ãããåã£ã¦ãã®ã§æ³¨æ
å¦ç¿çµæãä¿åã§ããããã«ãªã£ããããã¥ã¼ã©ã«ãããã®æ§æãJSONã§è¨è¿°ã§ããããã«ãªã£ãããã ã㶠ã¾ã£ã¨ããªãã¬ã¼ã ã¯ã¼ã¯ã«ãªã£ã¦ããæ°ããã¾ãã