Sign In

Twice Sequential Monte Carlo for Tree Search

์ž‘์„ฑ์ž
  • Haebom
์นดํ…Œ๊ณ ๋ฆฌ
Empty

์ €์ž

Yaniv Oren, Joery A. de Vries, Pascal R. van der Vaart, Matthijs T. J. Spaan, Wendelin Bohmer

๐Ÿ’ก ๊ฐœ์š”

๋ณธ ๋…ผ๋ฌธ์€ ๋ชจ๋ธ ๊ธฐ๋ฐ˜ ๊ฐ•ํ™”ํ•™์Šต์—์„œ ๋„๋ฆฌ ์‚ฌ์šฉ๋˜๋Š” MCTS์˜ ๋Œ€์•ˆ์œผ๋กœ ๋“ฑ์žฅํ•œ SMC ๋ฐฉ๋ฒ•๋ก ์ด ๊ฒช๋Š” ๋ถ„์‚ฐ ๋ฌธ์ œ์™€ ๊ฒฝ๋กœ ํ‡ดํ™” ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•ด TSMCTS(Twice Sequential Monte Carlo Tree Search)๋ฅผ ์ œ์•ˆํ•ฉ๋‹ˆ๋‹ค. TSMCTS๋Š” ๋””์Šคํฌ๋ฆฌํŠธ ๋ฐ ์—ฐ์† ํ™˜๊ฒฝ ๋ชจ๋‘์—์„œ ๊ธฐ์กด SMC ๋ฐ MCTS ๊ธฐ๋ฐ˜ ์ •์ฑ… ๊ฐœ์„  ๋ฐฉ๋ฒ•๋ณด๋‹ค ๋›ฐ์–ด๋‚œ ์„ฑ๋Šฅ์„ ๋ณด์ด๋ฉฐ, ์ˆœ์ฐจ์  ์ปดํ“จํŒ… ์ž์› ์ฆ๊ฐ€์— ๋”ฐ๋ผ ํ™•์žฅ์„ฑ์ด ์šฐ์ˆ˜ํ•˜๊ณ  ๋ถ„์‚ฐ์„ ์ค„์ด๋ฉฐ ๊ฒฝ๋กœ ํ‡ดํ™” ๋ฌธ์ œ๋ฅผ ์™„ํ™”ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ”‘ ์‹œ์‚ฌ์  ๋ฐ ํ•œ๊ณ„

โ€ข
TSMCTS๋Š” SMC์˜ ๋ณ‘๋ ฌํ™” ์šฉ์ด์„ฑ์„ ์œ ์ง€ํ•˜๋ฉด์„œ๋„ ๋ถ„์‚ฐ๊ณผ ๊ฒฝ๋กœ ํ‡ดํ™” ๋ฌธ์ œ๋ฅผ ํšจ๊ณผ์ ์œผ๋กœ ํ•ด๊ฒฐํ•˜์—ฌ ๋ณต์žกํ•œ ํ™˜๊ฒฝ์—์„œ์˜ ๊ฐ•ํ™”ํ•™์Šต ์„ฑ๋Šฅ์„ ํ–ฅ์ƒ์‹œํ‚ต๋‹ˆ๋‹ค.
โ€ข
์ˆœ์ฐจ์  ์ปดํ“จํŒ… ์ž์› ์ฆ๊ฐ€์— ๋Œ€ํ•œ ํ™•์žฅ์„ฑ์ด ์šฐ์ˆ˜ํ•˜์—ฌ ๊นŠ์€ ํƒ์ƒ‰์ด ํ•„์š”ํ•œ ๋ฌธ์ œ์— ๋”์šฑ ์ ํ•ฉํ•ฉ๋‹ˆ๋‹ค.
โ€ข
TSMCTS์˜ ์ด๋ก ์  ๋ถ„์„์ด๋‚˜ ์‹ค์ œ ์ ์šฉ ์‹œ ๋ฐœ์ƒํ•  ์ˆ˜ ์žˆ๋Š” ์ถ”๊ฐ€์ ์ธ ํ•˜์ดํผํŒŒ๋ผ๋ฏธํ„ฐ ํŠœ๋‹์˜ ํ•„์š”์„ฑ์— ๋Œ€ํ•œ ์—ฐ๊ตฌ๊ฐ€ ๋” ํ•„์š”ํ•ฉ๋‹ˆ๋‹ค.
๐Ÿ‘