AlphaGo 㨠AlphaGo Zero ã®è‡ªå·±å¯¾æˆ¦ã«ã‚ˆã‚‹å¦ç¿’部分ã®é•ã„ - ã†ã©ã‚“記æµã—èªã¿ã ã¨ã¡ã‚ƒã‚“ã¨åˆ†ã‹ã‚‰ãªã‹ã£ãŸã®ã§ãƒ¡ãƒ¢ã€‚ 準備(AlphaGo) policy network : 盤é¢ã¨ãã®ç‰¹å¾´é‡ã‚’入力ã¨ã—ã¦å—ã‘å–ã‚Šã€å„マスã«æ‰“ã¤ç¢ºçŽ‡ã‚’è¿”ã™ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã€‚ value network: 盤é¢ã¨ãã®ç‰¹å¾´é‡ã‚’入力ã¨ã—ã¦å—ã‘å–ã‚Šã€ãã®ç›¤é¢ã§ã®å‹çŽ‡ã‚’è¿”ã™ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã€‚ AlphaGo ã§ã¯ã¾ãš policy network をプãƒã®æ£‹èœãƒ‡ãƒ¼ã‚¿ã‹ã‚‰æ•™å¸«ã‚ã‚Šå¦ç¿’ã§äº‹å‰å¦ç¿’ã•ã›ã€ãã®å¾Œè‡ªå·±å¯¾æˆ¦ã«ã‚ˆã‚‹å¼·åŒ–å¦ç¿’ã«ã‚ˆã£ã¦ã•ã‚‰ã«æ”¹å–„ã•ã›ã¦ã„ã。 AlphaGo ã®å¼·åŒ–å¦ç¿’パート 教師ã‚ã‚Šå¦ç¿’後㮠policy network ã®ãƒ‘ラメータ $\rho_0$ ã‹ã‚‰å¦ç¿’をスタートã™ã‚‹ã€‚自己対戦ã®çµæžœã‹ã‚‰ policy network ã®ãƒ‘ラメータã¯éšæ™‚æ›´æ–°ã•ã‚Œã¦ã„ã。ãれらを $\rho_1, \rho_2, \cdots$ ã¨ã™ã‚‹ã€‚$t$ 回目ã®è‡ªå·±å¯¾æˆ¦ã§ã¯ã€ç¾åœ¨ã®ãƒ‘ラメータ $\
LIMEã§æ©Ÿæ¢°å¦ç¿’ã®äºˆæ¸¬çµæžœã‚’解釈ã—ã¦ã¿ã‚‹ - Qiita
Zero-shot / One-shot Learningã«é–¢ã™ã‚‹è³‡æ–™é›† - ã‚ã‚‚
ICML2017 å‚åŠ å ±å‘Š | CyberAgent Developers Blog
"SLIM: Sparse Linear Methods for Top-N Recommender Systems"ã‚’èªã‚“ã 
機械å¦ç¿’ã§å¸¸è˜ã‚’ç²å¾—ã™ã‚‹è«–文をèªã‚€ - mabonki0725ã®æ—¥è¨˜
「機械å¦ç¿’ï¼šæŠ€è¡“çš„è² å‚µã®é«˜åˆ©åクレジットカードã€ã®ã¾ã¨ã‚ | Recruit Tech Blog
ランプæ失サãƒãƒ¼ãƒˆãƒ™ã‚¯ã‚¿â€•ãƒžã‚·ãƒ³ - 何らã‹ã®ãƒ–ãƒã‚°
Yahoo!ã®ç•°å¸¸æ¤œçŸ¥ãƒ•ãƒ¬ãƒ¼ãƒ ワーク"EGADS"
GitHub - terryum/awesome-deep-learning-papers: The most cited deep learning papers
A Latent Variable Model Approach to PMI-based Word Embeddings
paragraph2vecã®è«–æ–‡ è¦è¨³ (Distributed Representations of Sentences and Documents) - Qiita
NIPS2013èªã¿ä¼š: Distributed Representations of Words and Phrases and their Compositionality
NIPS 2016 輪èª: Supervised Word Movers Distance
Word Mover's Distance: word2vecã®æ–‡æ›¸é–“è·é›¢ã¸ã®å¿œç”¨ - yubessy.hatenablog.com
主è¦ç§‘å¦èªŒã§ グーグルã«ã‚ˆã‚‹AIç ”ç©¶ã® è«–æ–‡ç™ºè¡¨ãŒæ€¥å¢—
introduce "Stealing Machine Learning Models via Prediction APIs"
機械å¦ç¿’界隈ã®æƒ…å ±åŽé›†æ–¹æ³• - kisa12012ã®æ—¥è¨˜
{{#tags}}- {{label}}
{{/tags}}