[B! å¼·åŒ–å¦ç¿’] yahihiã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

yahihi id:yahihi

å¼·åŒ–å¦ç¿’ã«é–¢ã™ã‚‹yahihiã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (5)

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

${{author_name}}$
{{author_name}}{{created}}
{{ #comment }}{{ comment }}{{ /comment }}
- {{ label }}

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

å¼·åŒ–å¦ç¿’æœªçµŒé¨“è€…ãŒãƒ†ãƒˆãƒªã‚¹ã® AI ã‚’ä½œã£ã¦ã¿ãŸè©±
ã¯ã˜ã‚ã« ã“ã®åº¦ã€å¼·åŒ–å¦ç¿’ã«ã‚ˆã‚‹ãƒ†ãƒˆãƒªã‚¹ã® AI ã‚’ä½œæˆã—ã¦ã¿ã¾ã—ãŸã€‚æƒ³åƒä»¥ä¸Šã«ã†ã¾ãã„ãã€æœ€çµ‚çš„ã«ã¯åŠæ°¸ä¹…çš„ã«ãƒ©ã‚¤ãƒ³ã‚’æ¶ˆã—ã¦ãã‚Œã‚‹ AI ã«æˆé•·ã—ã¦ãã‚ŒãŸã®ã§ã€ä»Šå›žã¯ãã®è¨˜éŒ²ã¨ã—ã¦ AI ã‚’ä½œæˆã—ãŸéŽç¨‹ã‚’ã“ã“ã«è¨˜ã—ã¦ã„ãã¾ã™ï¼ ä»Šå›žã® AI ã®æ¦‚è¦ ä»Šå›žä½œæˆã—ãŸãƒ†ãƒˆãƒªã‚¹ AI ã®æ¦‚è¦ã¯ä»¥ä¸‹ã®é€šã‚Šã§ã™ã€‚ ç‰¹å®šã®ã‚²ãƒ¼ãƒ ã®çŠ¶æ…‹ã‹ã‚‰å¯èƒ½ãªè¡Œå‹•ãƒ‘ã‚¿ãƒ¼ãƒ³ã‚’å…¨ã¦ã‚·ãƒŸãƒ¥ãƒ¬ãƒ¼ã‚·ãƒ§ãƒ³ã™ã‚‹ è¡Œå‹•å¾Œã®çŠ¶æ…‹ã‚’å…¥åŠ›æƒ…å ±ã¨ã—ã¦ Neural Network ã«ä»Šå¾Œã®å ±é…¬ã®æœŸå¾…å€¤ã‚’æŽ¨æ¸¬ã•ã›ã‚‹ ä»Šå¾Œã®å ±é…¬ã®æœŸå¾…å€¤ãŒä¸€ç•ªé«˜ããªã‚‹è¡Œå‹•ã‚’å®Ÿéš›ã®è¡Œå‹•ã¨ã—ã¦é¸æŠžã—ã¦é·ç§»ã™ã‚‹ (è²ªæ¬²æ–¹ç–) Game Over ã«ãªã‚‹ã¾ã§ 1 ~ 3 ã‚’ç¹°ã‚Šè¿”ã™ Tetris AI ã‚’å¯è¦–åŒ–ã—ãŸå›³ ä½•ã‚‚çŸ¥ã‚‰ãªã„é ƒã¯ã€ŽAI ã£ã¦ã©ã†ã‚„ã£ã¦å‹•ã„ã¦ã‚‹ã‚“ã ï¼Ÿã€ã¨è‡ªåˆ†ã¯æ€ã£ã¦ã„ãŸã‚“ã§ã™ãŒã€å‹•ä½œåŽŸç†ã¯éžå¸¸ã«ã‚·ãƒ³ãƒ—ãƒ«ã§ã™ã€‚å¼·åŒ–å¦ç¿’ã¨ã¯ã€ã“ã®ä»Šå¾Œã®å ±é…¬ã®æœŸå¾…å€¤ã‚’æŽ¨æ¸¬
yahihi 2024/05/01
å¼·åŒ–å¦ç¿’
ãƒªãƒ³ã‚¯
keras ä½¿ã£ã¦ DQN ã§è¿·è·¯ã‚’è§£ã„ã¦ã¿ãŸ - mizchi's blog
ä¸–ç•Œè¦³ã‚’ã¤ã‹ã‚ã‚‹ãã‚‰ã„ã«ã¯æ©Ÿæ¢°å¦ç¿’ã‚„ã£ã¦ãŠããŸã„ã¨æ€ã„ã€ã¨ã‚Šã‚ãˆãšä½•ã‹ã—ã‚‰ã®ãŠé¡ŒãŒãªã„ã¨èˆˆå‘³ãŒç¶šã‹ãªã•ãã†ãªã®ã§ã€äºŒæ¬¡å…ƒã®ç›¤é¢ä¸Šã§ä½•ã‹ã—ã‚‰ã®è¡Œå‹•ã‚’ã™ã‚‹ã€ãƒãƒ¼ã‚°ãƒ©ã‚¤ã‚¯ã®ãƒ¢ãƒ³ã‚¹ã‚¿ãƒ¼ã®ã‚¨ãƒ¼ã‚¸ã‚§ãƒ³ãƒˆã‚’ä½œã‚‹ã®ã‚’ç›®æ¨™ã«ã—ã‚ˆã†ã¨æ€ã†ã€‚è‡ªåˆ†ãŒã‚²ãƒ¼ãƒ ä½œã‚‹ã¨ãã€å¤§æŠµã‚¨ãƒ¼ã‚¸ã‚§ãƒ³ãƒˆã®ãƒ«ãƒ¼ãƒ«ä½œã‚‹æ®µéšŽã§é£½ãã¦ãã‚‹ã®ã§ã€‚ ä»Šå›žã®è¨˜äº‹ã¯ã€è¿·è·¯ã‚’è§£ãã¨ã“ã‚ã¾ã§ã€‚ å¦ç¿’è³‡æ–™ [Python]å¼·åŒ–å¦ç¿’(DQN)ã‚’å®Ÿè£…ã—ãªãŒã‚‰Kerasã«æ…£ã‚Œã‚‹ - Qiita DQNã‚’Kerasã¨TensorFlowã¨OpenAI Gymã§å®Ÿè£…ã™ã‚‹ å…¨åŠ›ã§äººå·¥çŸ¥èƒ½ã«å¯¾æ±ºã‚’æŒ‘ã‚“ã§ã¿ãŸï¼ˆç†è«–ç·¨ï¼‰ - ãƒ‹ã‚³ãƒ‹ã‚³å‹•ç”» é›°å›²æ°—ã‚’æŽ´ã‚€ã®ã«ã€ãƒ‹ã‚³å‹•ã®è§£èª¬å‹•ç”»ã‚ã‹ã‚Šã‚„ã™ã‹ã£ãŸã€‚ ã‚ˆãä½¿ã‚ã‚Œã¦ã‚‹ OpenAI Gym ã€è¦‹ãŸç›®ã¯æ´¾æ‰‹ã ãŒã€ç’°å¢ƒå¤‰æ•°ãŒå¤šã™ãŽã¦ã„ã¾ã„ã¡ç†è§£ã®åŠ©ã‘ã«ãªã‚‰ãªã„ + æ¬¡å…ƒãŒå¤šã™ãŽã¦åŽæŸãŒé ã„ã®ã§ã€ã™ã”ã„å˜ç´”ãªã‚²ãƒ¼ãƒ ãƒ«ãƒ¼ãƒ«ã‚’è‡ªåˆ†ã§ä½œã£ã¦ã€ãã‚Œ
yahihi 2018/10/29
python

å¼·åŒ–å¦ç¿’
ãƒªãƒ³ã‚¯
Googleã€TensorFlowãƒ™ãƒ¼ã‚¹ã®å¼·åŒ–å¦ç¿’ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ã€ŒDopamineã€ã‚’å…¬é–‹ | OSDN Magazine
Googleã¯8æœˆ27æ—¥ã€å¼·åŒ–å¦ç¿’ç ”ç©¶ã®ãŸã‚ã®ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ã€ŒDopamineã€ã‚’ã‚ªãƒ¼ãƒ—ãƒ³ã‚½ãƒ¼ã‚¹ã§å…¬é–‹ã—ãŸã€‚å¼·åŒ–å¦ç¿’ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã®ãƒ—ãƒãƒˆã‚¿ã‚¤ãƒ—ã‚’é«˜é€Ÿã«å®Ÿè£…ã§ãã‚‹ã¨ã„ã†ã€‚ å¼·åŒ–å¦ç¿’ã¯å…¥åŠ›ã«å¯¾ã™ã‚‹ã€Œå ±é…¬ã€ãŒæœ€å¤§ã«ãªã‚‹ã‚ˆã†ãªè¡Œå‹•ã‚’å¦ç¿’ã«ã‚ˆã£ã¦æ±ºå®šã™ã‚‹æ‰‹æ³•ã€‚äººé–“ã®å›²ç¢æ£‹å£«ã«å‹åˆ©ã—ãŸGoogle DeepMindã®å›²ç¢ã‚½ãƒ•ãƒˆã‚¦ã‚§ã‚¢ã€ŒAlpha Goã€ãªã©ã§æŽ¡ç”¨ã•ã‚Œã¦ã„ã‚‹æ©Ÿæ¢°å¦ç¿’æ‰‹æ³•ã®1ã¤ã€‚ Googleã«ã‚ˆã‚‹ã¨ã€å¼·åŒ–å¦ç¿’ã®èª²é¡Œã¨ã—ã¦æ—¢å˜ã®å¼·åŒ–å¦ç¿’ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ã¯ç ”ç©¶è€…ãŒåŠ¹æžœçš„ã«ã‚¤ãƒ†ãƒ¬ãƒ¼ã‚·ãƒ§ãƒ³ã™ã‚‹ã®ã«ååˆ†ãªæŸ”è»Ÿæ€§ã¨å®‰å…¨æ€§ãŒãªãã€æ–°ã—ã„ç ”ç©¶ã®æ–¹å‘æ€§ã‚’æŽ¢ã‚‹ã“ã¨ãŒé›£ã—ã„ã¨ã„ã†å•é¡ŒãŒã‚ã‚‹ã¨ã„ã†ã€‚ã¾ãŸã€æ—¢å˜ã®ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ã‹ã‚‰ã®çµæžœã®å†ç¾ã‚‚æ™‚é–“ãŒã‹ã‹ã‚‹ã¨ã„ã†ã€‚ Dopamineã¯æ©Ÿæ¢°å¦ç¿’ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã®TensorFlowã‚’åœŸå°ã¨ã—ãŸã‚ªãƒ¼ãƒ—ãƒ³ã‚½ãƒ¼ã‚¹ã®å¼·åŒ–å¦ç¿’ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ã€‚ãƒ™ãƒ³ãƒãƒžãƒ¼ã‚¯ãƒ†ã‚¹ãƒˆã‚’ç°¡å˜ã«å®Ÿè¡Œã§ãã‚‹ã€æ–°ã—
yahihi 2018/08/30
tensorflow

google

å¼·åŒ–å¦ç¿’
ãƒªãƒ³ã‚¯
GitHub - google/dopamine: Dopamine is a research framework for fast prototyping of reinforcement learning algorithms.
Dopamine is a research framework for fast prototyping of reinforcement learning algorithms. It aims to fill the need for a small, easily grokked codebase in which users can freely experiment with wild ideas (speculative research). Our design principles are: Easy experimentation: Make it easy for new users to run benchmark experiments. Flexible development: Make it easy for new users to try out res
yahihi 2018/08/29
google

python

å¼·åŒ–å¦ç¿’
ãƒªãƒ³ã‚¯
å¼·åŒ–å¦ç¿’å…¥é–€ Part3 ï¼ AlphaGoZeroã§ã‚‚é‡è¦ãªæŠ€è¡“è¦ç´ ï¼ ãƒ¢ãƒ³ãƒ†ã‚«ãƒ«ãƒæœ¨æŽ¢ç´¢ã®å…¥é–€ ï¼ - Platinum Data Blog by BrainPad
æœ¬è¨˜äº‹ã¯ã€å½“ç¤¾ã‚ªã‚¦ãƒ³ãƒ‰ãƒ¡ãƒ‡ã‚£ã‚¢ã€ŒDoorsã€ã«ç§»è»¢ã—ã¾ã—ãŸã€‚ ç´„5ç§’å¾Œã«è‡ªå‹•çš„ã«ãƒªãƒ€ã‚¤ãƒ¬ã‚¯ãƒˆã—ã¾ã™ã€‚ å¼·åŒ–å¦ç¿’å…¥é–€ã®ç¬¬3å¼¾ã€‚ã€Œãƒ¢ãƒ³ãƒ†ã‚«ãƒ«ãƒæœ¨æŽ¢ç´¢ï¼ˆMonte Carlo Tree Search, MCTSï¼‰ã€ã‚’è§£èª¬ã™ã‚‹ã¨ã¨ã‚‚ã«ã€å®Ÿéš›ã«AIåŒå£«ã§äº”ç›®ä¸¦ã¹ã‚’æˆ¦ã‚ã›ã¦ã¿ã¾ã—ãŸï¼ ã“ã‚“ã«ã¡ã¯ã€‚ã‚¢ãƒŠãƒªãƒ†ã‚£ã‚¯ã‚¹ã‚µãƒ¼ãƒ“ã‚¹æœ¬éƒ¨ AIé–‹ç™ºéƒ¨ã®å±±å´Žã§ã™ã€‚ æ˜¨å¹´ã‚‚å¼·åŒ–å¦ç¿’ç•Œéšˆã¯ç››ã‚Šä¸ŠãŒã‚Šã‚’è¦‹ã›ã¦ã„ã¾ã—ãŸãŒã€ä»ŠãªãŠã€å¼·åŒ–å¦ç¿’ã¨è¨€ãˆã°ä¸€ç•ªã«Alpha Goã‚’æ€ã„æµ®ã‹ã¹ã‚‹äººã‚‚å¤šã„ã®ã§ã¯ãªã„ã§ã—ã‚‡ã†ã‹ã€‚æ˜¨å¹´ã€Alpha GoZeroã¨ã„ã†é€²åŒ–ãƒãƒ¼ã‚¸ãƒ§ãƒ³ãŒç™ºè¡¨ã•ã‚Œã€ä¸€åˆ‡ã®ãŠæ‰‹æœ¬ã‚’ç”¨ã„ãšã«å¾“æ¥ãƒãƒ¼ã‚¸ãƒ§ãƒ³ã‚ˆã‚Šå¼·ã„ã¨ã„ã†ã“ã¨ãŒè©±é¡Œã«ãªã‚Šã¾ã—ãŸã€‚ï¼ˆãã‚Œã¾ã§ã®Alpha Goã¯ã€äººã®ãŠæ‰‹æœ¬ã‚’ã‚ã‚‹ç¨‹åº¦å¦ã³ã€ãã®å¾Œå‹æ‰‹ã«å¦ã‚“ã§ã„ãã‚‚ã®ã§ã—ãŸï¼‰ ã•ã‚‰ã«ã€Alpha GoZeroã‚’ä¸€èˆ¬åŒ–ã—ãŸã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã«ä¿®æ£ã—ãŸAlphaZeroãŒç™»å ´ã—ã€ãƒã‚§ã‚¹
yahihi 2018/05/08
ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ

AlphaGo

å¼·åŒ–å¦ç¿’

ãƒ¢ãƒ³ãƒ†ã‚«ãƒ«ãƒæœ¨
ãƒªãƒ³ã‚¯
1