Sign In

AstRL: Analog and Mixed-Signal Circuit Synthesis with Deep Reinforcement Learning

Created by
  • Haebom
Category
Empty

์ €์ž

Felicia B. Guo, Ken T. Ho, Andrei Vladimirescu, Borivoje Nikolic

๐Ÿ’ก ๊ฐœ์š”

๋ณธ ๋…ผ๋ฌธ์€ ๋ณต์žกํ•˜๊ณ  ๋น„์„ ํ˜•์ ์ธ ์•„๋‚ ๋กœ๊ทธ ๋ฐ ํ˜ผํ•ฉ ์‹ ํ˜ธ(AMS) ํšŒ๋กœ ์„ค๊ณ„ ์ž๋™ํ™”์˜ ์–ด๋ ค์›€์„ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•ด, ํšŒ๋กœ ์„ค๊ณ„๋ฅผ ๊ทธ๋ž˜ํ”„ ์ƒ์„ฑ ๋ฌธ์ œ๋กœ ์ ‘๊ทผํ•˜๊ณ  ๋”ฅ ๊ฐ•ํ™”ํ•™์Šต ๊ธฐ๋ฐ˜์˜ ์ƒˆ๋กœ์šด ํ•ฉ์„ฑ ๋ฐฉ๋ฒ•๋ก ์ธ AstRL์„ ์ œ์•ˆํ•œ๋‹ค. AstRL์€ ์ •์ฑ… ๊ฒฝ์‚ฌ ํ•™์Šต ์ ‘๊ทผ๋ฒ•์„ ๊ธฐ๋ฐ˜์œผ๋กœ ์‹œ๋ฎฌ๋ ˆ์ดํ„ฐ ํ™˜๊ฒฝ์—์„œ ์‹ค์ œ์ ์ธ ํ”ผ๋“œ๋ฐฑ์„ ๋ฐ›์•„ ์‚ฌ์šฉ์ž ์ง€์ • ๋ชฉํ‘œ์— ์ตœ์ ํ™”๋œ ํšŒ๋กœ๋ฅผ ์ง์ ‘ ์ƒ์„ฑํ•˜๋ฉฐ, ํ–‰๋™ ๋ณต์ œ ๋ฐ ํŒ๋ณ„๊ธฐ ๊ธฐ๋ฐ˜ ์œ ์‚ฌ๋„ ๋ณด์ƒ์„ ํ†ตํ•ด ์ „๋ฌธ๊ฐ€ ์ˆ˜์ค€์˜ ํšŒ๋กœ ์ƒ์„ฑ์„ ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ํ™˜๊ฒฝ์—์„œ ์ตœ์ดˆ๋กœ ๊ฒ€์ฆํ•˜์˜€๋‹ค. ํŠนํžˆ, ๊ฐœ๋ณ„ ํŠธ๋žœ์ง€์Šคํ„ฐ ์ˆ˜์ค€์—์„œ ์ž‘๋™ํ•˜์—ฌ ๋ฏธ์„ธํ•˜๊ณ  ํ‘œํ˜„๋ ฅ์ด ํ’๋ถ€ํ•œ ํ† ํด๋กœ์ง€ ์ƒ์„ฑ์ด ๊ฐ€๋Šฅํ•˜๋ฉฐ, ์‹คํ—˜ ๊ฒฐ๊ณผ๋Š” ๊ธฐ์กด ์„ค๊ณ„ ๋ฐฉ์‹ ๋Œ€๋น„ ์ƒ๋‹นํ•œ ์„ฑ๋Šฅ ํ–ฅ์ƒ์„ ๋ณด์—ฌ์ฃผ์—ˆ๋‹ค.

๐Ÿ”‘ ์‹œ์‚ฌ์  ๋ฐ ํ•œ๊ณ„

โ€ข
๋”ฅ ๊ฐ•ํ™”ํ•™์Šต์„ ํ™œ์šฉํ•˜์—ฌ ๋ณต์žกํ•˜๊ณ  ๋น„์„ ํ˜•์ ์ธ AMS ํšŒ๋กœ ์„ค๊ณ„ ์ž๋™ํ™”์˜ ๊ฐ€๋Šฅ์„ฑ์„ ์ œ์‹œํ•˜์˜€๋‹ค.
โ€ข
๊ฐœ๋ณ„ ํŠธ๋žœ์ง€์Šคํ„ฐ ์ˆ˜์ค€์—์„œ์˜ ์ƒ์„ธํ•œ ํ† ํด๋กœ์ง€ ์ƒ์„ฑ์„ ํ†ตํ•ด ๊ธฐ์กด ๋ฐฉ๋ฒ•๋ก  ๋Œ€๋น„ ๋›ฐ์–ด๋‚œ ์„ฑ๋Šฅ์„ ๋‹ฌ์„ฑํ•˜์˜€๋‹ค.
โ€ข
์ƒ์„ฑ๋œ ํšŒ๋กœ์˜ 100% ๊ตฌ์กฐ์  ์ •ํ™•์„ฑ๊ณผ 90% ์ด์ƒ์˜ ๊ธฐ๋Šฅ์„ฑ ํ™•๋ณด๋Š” ์‹ค์ œ ์„ค๊ณ„ ์ ์šฉ ๊ฐ€๋Šฅ์„ฑ์„ ๋†’์ธ๋‹ค.
โ€ข
์‹ค์ œ ํ•˜๋“œ์›จ์–ด ๊ตฌํ˜„์—์„œ์˜ ๊ฒ€์ฆ ๋ฐ ๋‹ค์–‘ํ•œ ์ข…๋ฅ˜์˜ AMS ํšŒ๋กœ ์„ค๊ณ„์— ๋Œ€ํ•œ ์ผ๋ฐ˜ํ™” ์„ฑ๋Šฅ ํ–ฅ์ƒ์ด ํ–ฅํ›„ ๊ณผ์ œ๋กœ ๋‚จ๋Š”๋‹ค.
๐Ÿ‘