ゼãƒã‹ã‚‰æ·±å±¤å¦ç¿’ã‚’å¦ã¶æ–¹æ³• - CMS大阪å¤ç¥ã‚Š20172. NTTコミュニケーションズ 増田 知彰 | @tomo_makes 本日ã®å†…容ã€ç™ºè¨€ã¯ã™ã¹ã¦å€‹äººã¨ã—ã¦ã®æ´»å‹•ã§ã‚ã‚Šã€ä¼šç¤¾ã‚’代表ã™ã‚‹ã‚‚ã®ã§ã¯ã‚ã‚Šã¾ã›ã‚“ 自己紹介 オン: IoT関連ã®BizDev Product Management オフ: Weekend Maker/Programmer データ分æž/機械å¦ç¿’ OSS活動(éžã‚³ãƒ¼ãƒ‡ã‚£ãƒ³ã‚°ä¸»ä½“) 6. 精度 4%ã‹ã‚‰78%ã«ã„ãŸã‚‹2ヶ月 DAY1 10 20 30 40 50 60 予測精度 å±¥æ´ 1/27 IDCF 貸出 â— 37.6% ç°¡å˜ãªCNNモデル â— 59.6% VGG16 fine-tuning â— 73.1% InceptionV3 fine-tuning 6 models blending â— 78.9% [暫定3ä½] â— 4% エラー ブãƒã‚°, arxivè«–æ–‡ MOOCs fast.ai part I Kaggle
fast.aiã® Cutting Edge Deep Learning For Coders, Part 2ã‚’å—ã‘ãŸ
FADL2 L8: ‘Neural’ Artistic Style Transfer
A simple pseudo-labeling implementation in keras
Running Fast.ai Lesson 1 on Floyd [Solved] - How To - FloydHub Forum
Running fast.ai course notebook on Floyd (Floydhub)
Today I Learned This Part 2: Pretrained Neural Networks What are they? - MLWhiz
{{#tags}}- {{label}}
{{/tags}}