KOINEU Logo
No Image

The Role of Common Randomness Replication in Symmetric PIR on Graph-Based Replicated Systems

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ PIR vs. SPIR : ์ „ํ†ต์ ์ธ PIR์€ ์‚ฌ์šฉ์ž ํ”„๋ผ์ด๋ฒ„์‹œ๋งŒ์„ ๋ณด์žฅํ•˜๊ณ , ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค ๋‚ด์šฉ์ด ์‚ฌ์šฉ์ž์—๊ฒŒ ์ผ๋ถ€ ๋…ธ์ถœ๋˜๋Š” ๊ฒƒ์ด ํ—ˆ์šฉ๋œ๋‹ค. ๋ฐ˜๋ฉด SPIR์€ ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค ํ”„๋ผ์ด๋ฒ„์‹œ๊นŒ์ง€ ์š”๊ตฌํ•˜๋ฏ€๋กœ, ์„œ๋ฒ„ ๊ฐ„์— ๊ณตํ†ต ๋žœ๋ค์„ฑ ์ด ํ•„์ˆ˜์ ์ด๋‹ค. ๊ทธ๋ž˜ํ”„ ๊ธฐ๋ฐ˜ ๋ณต์ œ : ์‹ค์ œ ํด๋ผ์šฐ๋“œยท๋ถ„์‚ฐ ์ €์žฅ ์‹œ์Šคํ…œ์—์„œ๋Š” ๋ชจ๋“  ๋ฉ”์‹œ์ง€๋ฅผ ๋ชจ๋“  ์„œ๋ฒ„์— ๋ณต์ œํ•˜๊ธฐ ์–ด๋ ค์›Œ, ๋‘ ์„œ๋ฒ„์—๋งŒ ๋ณต์ œ ๋˜๋Š” ๊ตฌ์กฐ๊ฐ€ ํ˜„์‹ค์ ์ด๋‹ค. ์ด๋ฅผ ๊ทธ๋ž˜ํ”„(์ •์  ์„œ๋ฒ„, ๊ฐ„์„  ๋ฉ”์‹œ์ง€)๋กœ ๋ชจ๋ธ๋งํ•˜๋ฉด ๋‹ค์–‘ํ•œ ๋„คํŠธ์›Œํฌ ํ† ํด๋กœ์ง€๋ฅผ ์ž์—ฐ์Šค๋Ÿฝ๊ฒŒ ํฌ์ฐฉํ•  ์ˆ˜ ์žˆ๋‹ค. ๋žœ๋ค์„ฑ ๋ฐฐํฌ์˜ ๋‘ ์ถ• :

Computer Science System Information Theory
No Image

The role of VSG parameters in shaping small-signal SG dynamics

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ์ „๋ ฅ ์ „์ž ์ธํ„ฐํŽ˜์ด์Šค์™€ ์žฌ์ƒ์—๋„ˆ์ง€ ํ™•๋Œ€ ๋กœ ์ธํ•ด ๊ทธ๋ฆฌ๋“œโ€‘ํฌ๋ฐ(GFM) ์ œ์–ด๊ฐ€ ํ•ต์‹ฌ ๊ธฐ์ˆ ๋กœ ๋ถ€์ƒํ•˜๊ณ  ์žˆ๋‹ค. ๊ธฐ์กด ์†Œ์‹ ํ˜ธ ๋ถ„์„์€ ๋‹จ์ผ ์ปจ๋ฒ„ํ„ฐโ€‘๊ทธ๋ฆฌ๋“œ ์—ฐ๊ฒฐ ์— ์ดˆ์ ์„ ๋งž์ถ”์—ˆ์œผ๋ฉฐ, VSG์™€ SG๊ฐ€ ๋ณ‘๋ ฌ๋กœ ์—ฐ๊ฒฐ๋œ ๊ฒฝ์šฐ์˜ ์ €์ฃผํŒŒยท๊ณ ์ฃผํŒŒ ์ƒํ˜ธ์ž‘์šฉ ์€ ์ถฉ๋ถ„ํžˆ ๋‹ค๋ฃจ์–ด์ง€์ง€ ์•Š์•˜๋‹ค. ๋ณธ ๋…ผ๋ฌธ์€ ๋„คํŠธ์›Œํฌยท์ „๋ฅ˜ ์ œ์–ด๋ฅผ ๋ฐฐ์ œํ•˜๊ณ  ๋А๋ฆฐ ๋™์—ญํ•™(๊ด€์„ฑยท๊ฐ€๋ฒ„๋„ˆยท์ „์••ยท๋ฌดํšจ์ „๋ ฅ)๋งŒ์„ ์ง‘์ค‘ ๋ถ„์„ํ•จ์œผ๋กœ์จ, ๋ณต์žกํ•œ ์ƒ์„ธ ๋ชจ๋ธ์ด ๊ฐ€๋ฆด ์ˆ˜ ์žˆ๋Š” ํ•ต์‹ฌ ๋ฉ”์ปค๋‹ˆ์ฆ˜์„ ๋“œ๋Ÿฌ๋‚ธ๋‹ค. 2. ๋ชจ๋ธ๋ง ์ ‘๊ทผ๋ฒ• | ํ•ญ๋ชฉ | ๋‚ด์šฉ | ํŠน์ง• | | | | | | ์‹œ์Šคํ…œ ๊ตฌ์„ฑ

Electrical Engineering and Systems Science
No Image

The Well-Tempered Classifier: Some Elementary Properties of Temperature Scaling

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ ๋ถˆํ™•์‹ค์„ฑ ์ •๋Ÿ‰ํ™” ๋Š” ํ˜„์žฌ ๋”ฅ๋Ÿฌ๋‹์—์„œ ๊ฐ€์žฅ ํ™œ๋ฐœํžˆ ๋…ผ์˜๋˜๋Š” ์ฃผ์ œ ์ค‘ ํ•˜๋‚˜์ด๋ฉฐ, ์ด๋ก ์  ํ•œ๊ณ„์™€ ์‹ค์šฉ์  ํ•ด๊ฒฐ์ฑ… ์‚ฌ์ด์— ํฐ ๊ฒฉ์ฐจ๊ฐ€ ์กด์žฌํ•œ๋‹ค(์˜ˆ: Foygelโ€‘Barber et al., 2021). ์˜จ๋„ ์Šค์ผ€์ผ๋ง์€ ๋‹จ์ผ ์Šค์นผ๋ผ ํŒŒ๋ผ๋ฏธํ„ฐ (ฮฒ)๋งŒ์œผ๋กœ ๋ชจ๋ธ์˜ ํ™•๋ฅ  ๋ถ„ํฌ๋ฅผ ์กฐ์ •ํ•  ์ˆ˜ ์žˆ์–ด, ์บ˜๋ฆฌ๋ธŒ๋ ˆ์ด์…˜๊ณผ LLM ๋””์ฝ”๋”ฉ ๋ชจ๋‘์—์„œ โ€œ๊ฐ€์žฅ ์‰ฌ์šดโ€ ๋ฐฉ๋ฒ•์œผ๋กœ ์ž๋ฆฌ ์žก์•˜๋‹ค. ๊ทธ๋Ÿผ์—๋„ ๋ถˆ๊ตฌํ•˜๊ณ  ์ด๋ก ์  ๋ถ„์„์ด ๋ถ€์กฑ ํ–ˆ์œผ๋ฉฐ, ๊ธฐ์กด ์—ฐ๊ตฌ๋Š” ์ฃผ๋กœ ๊ฒฝํ—˜์  ํ‰๊ฐ€์— ๋จธ๋ฌผ๋ €๋‹ค. ๋ณธ ๋…ผ๋ฌธ์€ ์ด ๊ณต๋ฐฑ์„ ๋ฉ”์šฐ๊ณ , ์˜จ๋„ ์Šค์ผ€์ผ๋ง์„ ์ˆ˜ํ•™์ ์œผ๋กœ ์—„๋ฐ€ํžˆ ์ดํ•ดํ•œ

Statistics Machine Learning
No Image

Thermal Decoherence and Population Transfer of MeV Channeling Electrons in Diamond

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ์ฑ„๋„๋ง ๋ณต์‚ฌ ๋Š” ์ „์ž๊ฐ€ ๊ฒฐ์ • ๋‚ด ์›์ž์—ด์— ์˜ํ•ด ํ˜•์„ฑ๋œ ์ฃผ๊ธฐ์  ํผํ…์…œ์— ์–ฝํžŒ ์–‘์žํ™”๋œ ํšก๋ฐฉํ–ฅ ์ƒํƒœ ์‚ฌ์ด ์ „์ดํ•˜๋ฉด์„œ ๋ฐœ์ƒํ•œ๋‹ค. ์ €์—๋„ˆ์ง€(MeV) ์˜์—ญ์—์„œ๋Š” ์ƒํƒœ๊ฐ€ ๋ช…ํ™•ํžˆ ์–‘์žํ™”๋˜๋ฏ€๋กœ, ์ „์ด์„  ์ŠคํŽ™ํŠธ๋Ÿผ์„ ์ •ํ™•ํžˆ ์˜ˆ์ธกํ•˜๋ ค๋ฉด ๋™์  ์—ดโ€‘์‚ฐ๋ž€ ์„ ํฌํ•จํ•œ ์ „์žโ€‘๊ฒฉ์ž ์ƒํ˜ธ์ž‘์šฉ์„ ๊ณ ๋ คํ•ด์•ผ ํ•œ๋‹ค. ๊ธฐ์กด ์ด๋ก ์€ Debyeโ€‘Waller ํ‰๊ท  ํผํ…์…œ ์„ ์ด์šฉํ•ด ์—๋„ˆ์ง€ ์ด๋™๋งŒ์„ ๋ณด์ •ํ–ˆ์œผ๋ฉฐ, ์ธ๊ตฌ ์ „์ด(population transfer) ์™€ ํƒˆ๋™์กฐ(decoherence) ๊ฐ™์€ ๋น„ํƒ„์„ฑ ํšจ๊ณผ๋Š” ๋ฌด์‹œ๋˜์—ˆ๋‹ค. ์ด๋Š” ์‹คํ—˜์—์„œ ๊ด€์ธก๋˜๋Š” ์ŠคํŽ™ํŠธ

Condensed Matter
No Image

Three dimensional contractile droplet under confinement

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ ์•กํ‹ฐ๋ธŒ ๋“œ๋กญ๋ฆฟ ์€ ๋‚ด๋ถ€ ํ˜น์€ ์ฃผ๋ณ€์—์„œ ์—๋„ˆ์ง€๋ฅผ ๊ณต๊ธ‰๋ฐ›์•„ ์ž๊ฐ€๊ตฌ๋™์„ ํ•˜๋Š” ์—ฐ์„ฑ ๋ฌผ์งˆ๋กœ, ์„ธํฌ ์ด๋™์„ฑยท์ธ๊ณต ๋งˆ์ดํฌ๋กœ์Šค์œ„๋จธ ์„ค๊ณ„์— ํ•ต์‹ฌ ๋ชจ๋ธ์ด๋‹ค. ๊ธฐ์กด ์—ฐ๊ตฌ๋Š” ์ฃผ๋กœ 2์ฐจ์› ๋˜๋Š” ๋ฌด์ œํ•œ(3D) ํ™˜๊ฒฝ์— ์ดˆ์ ์„ ๋งž์ถ”์—ˆ์œผ๋ฉฐ, ์ œํ•œ๋œ 3D ํ™˜๊ฒฝ ์—์„œ์˜ ๊ฑฐ๋™์€ ์•„์ง ์ถฉ๋ถ„ํžˆ ํƒ๊ตฌ๋˜์ง€ ์•Š์•˜๋‹ค. ๋ณธ ๋…ผ๋ฌธ์€ ์ด๋Ÿฌํ•œ ๊ณต๋ฐฑ์„ ๋ฉ”์šฐ๊ธฐ ์œ„ํ•ด 3์ฐจ์› ๋ฏธ์„ธ์ฑ„๋„ ๋‚ด์—์„œ ์ˆ˜์ถ•์„ฑ(active contractile) ๋“œ๋กญ๋ฆฟ์˜ ๋™์—ญํ•™์„ ์ฒด๊ณ„์ ์œผ๋กœ ๋ถ„์„ํ•œ๋‹ค. 2. ๋ชจ๋ธ๋ง ๋ฐ ์ˆ˜์น˜ ๋ฐฉ๋ฒ• | ์š”์†Œ | ์„ค๋ช… | ์ฃผ์š” ํŒŒ๋ผ๋ฏธํ„ฐ | | | | | | ์ƒ(pha

Condensed Matter
No Image

Tileable Surfaces

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ ํƒ€์ผ๋ง ์ด๋ก ์€ ํ‰๋ฉดยท๊ณก๋ฉดยท๋‹ค์–‘์ฒด ์ „๋ฐ˜์— ๊ฑธ์ณ ์กฐํ•ฉ๋ก ยท๊ธฐํ•˜ํ•™ยท๋ฌผ๋ฆฌํ•™ ๋“ฑ ๋‹ค์–‘ํ•œ ๋ถ„์•ผ์™€ ์—ฐ๊ฒฐ๋œ๋‹ค. ํŠนํžˆ ๋‹จ์ผ ํƒ€์ผ๋ง(monotiling) ์€ โ€œํ•˜๋‚˜์˜ ์›ํ˜•(ํ”„๋กœํ† ํƒ€์ผ)๋งŒ์œผ๋กœ ์ „์ฒด๋ฅผ ์ฑ„์šธ ์ˆ˜ ์žˆ๋‹คโ€๋Š” ์ œํ•œ๋œ ์กฐ๊ฑด ๋•Œ๋ฌธ์— ๋ณต์žกํ•œ ์œ„์ƒยท๊ธฐํ•˜ํ•™์  ๊ตฌ์กฐ๋ฅผ ๋“œ๋Ÿฌ๋‚ด๋Š” ์ข‹์€ ์‹คํ—˜๋Œ€๊ฐ€ ๋œ๋‹ค. ๊ธฐ์กด ์—ฐ๊ตฌ๋Š” ์ฃผ๋กœ ํ‰๋ฉดยท์ •๋‹ค๊ฐํ˜•์— ๋จธ๋ฌผ๋ €์œผ๋ฉฐ, 3์ฐจ์› ๊ณก๋ฉด์— ๋Œ€ํ•œ ์ฒด๊ณ„์ ์ธ ๋ถ„๋ฅ˜๋Š” ๊ฑฐ์˜ ์ด๋ฃจ์–ด์ง€์ง€ ์•Š์•˜๋‹ค. ๋ณธ ๋…ผ๋ฌธ์€ ์ด๋Ÿฌํ•œ ๊ณต๋ฐฑ์„ ๋ฉ”์šฐ๊ณ , (C^1) ์ •๊ทœ์„ฑ ์ด๋ผ๋Š” ์ตœ์†Œํ•œ์˜ ๋งค๋„๋Ÿฌ์›€ ์กฐ๊ฑด ํ•˜์—์„œ ํƒ€์ผ๋ง ๊ฐ€๋Šฅํ•œ ๊ณก๋ฉด์„ ํƒ๊ตฌํ•œ๋‹ค๋Š” ์ ์—์„œ

No Image

Time-Certified and Efficient NMPC via Koopman Operator

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ NMPC๋Š” ๋งค ์ƒ˜ํ”Œ๋ง ์‹œ์ ๋งˆ๋‹ค ์ตœ์ ํ™” ๋ฌธ์ œ๋ฅผ ํ’€์–ด์•ผ ํ•˜๋ฏ€๋กœ, ํ‰๊ท  ์‹คํ–‰ ์‹œ๊ฐ„ ๊ณผ ์ตœ์•…โ€‘์‚ฌ๋ก€ ์‹คํ–‰ ์‹œ๊ฐ„ ๋‘ ๊ฐ€์ง€ ์š”๊ตฌ์‚ฌํ•ญ์ด ๋™์‹œ์— ์ถฉ์กฑ๋ผ์•ผ ์‹ค์‹œ๊ฐ„ ์ ์šฉ์ด ๊ฐ€๋Šฅํ•˜๋‹ค. ๊ธฐ์กด ์—ฐ๊ตฌ๋“ค์€ ํ‰๊ท  ์‹คํ–‰ ์‹œ๊ฐ„ ๊ฐ์†Œ์— ์ง‘์ค‘ํ–ˆ์œผ๋‚˜, ์ตœ์•…โ€‘์‚ฌ๋ก€ ์‹คํ–‰ ์‹œ๊ฐ„์— ๋Œ€ํ•œ ๋ฐ์ดํ„ฐโ€‘๋…๋ฆฝ์ ์ธ ์ธ์ฆ ์€ ๊ฑฐ์˜ ์—†์—ˆ๋‹ค. ์ด๋Š” ํŠนํžˆ ์ž„๋ฒ ๋””๋“œ ์‹œ์Šคํ…œ์—์„œ ์•ˆ์ „์„ฑ์„ ๋ณด์žฅํ•˜๊ธฐ ์–ด๋ ต๊ฒŒ ๋งŒ๋“ ๋‹ค. 2. Koopman ์—ฐ์‚ฐ์ž๋ฅผ ํ†ตํ•œ ๋ชจ๋ธ ๊ณ ์ฐจ์› ์„ ํ˜•ํ™” Koopman ์—ฐ์‚ฐ์ž๋Š” ๋น„์„ ํ˜• ๋™์—ญํ•™์„ ๋ฌดํ•œ ์ฐจ์›์˜ ์„ ํ˜• ์—ฐ์‚ฐ์ž๋กœ ํ‘œํ˜„ํ•œ๋‹ค. ์‹ค์šฉ์ ์œผ๋กœ๋Š” EDMD (Extend

Electrical Engineering and Systems Science
No Image

Time-Varying Directed Interactions in Functional Brain Networks: Modeling and Validation

| ํ•ญ๋ชฉ | ๋‚ด์šฉ ๋ฐ ํ‰๊ฐ€ | | | | | ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ | ๊ธฐ์กด ๋™์  FC ๋ถ„์„์€ ์ฃผ๋กœ ์ƒ๊ด€ ๊ธฐ๋ฐ˜(SWC)์œผ๋กœ, ๋ฐฉํ–ฅ์„ฑ์„ ๋ฐ˜์˜ํ•˜์ง€ ๋ชปํ•œ๋‹ค. ์ธ๊ณผ ๋ชจ๋ธ๊ณผ ๋™์  ๋ถ„์„์„ ๊ฒฐํ•ฉํ•˜๋ ค๋Š” ์‹œ๋„๋Š” ๋“œ๋ฌผ๋ฉฐ, ํŠนํžˆ ์ „์—ญ ๋‡Œ ๋„คํŠธ์›Œํฌ์— ์ ์šฉํ•˜๊ธฐ ์–ด๋ ค์› ๋‹ค. ๋ณธ ๋…ผ๋ฌธ์€ ์ด ๊ฒฉ์ฐจ๋ฅผ ๋ฉ”์šฐ๊ธฐ ์œ„ํ•ด LTI ์ธ๊ณผ ๋ชจ๋ธ์„ ์Šฌ๋ผ์ด๋”ฉ ์œˆ๋„์šฐ์— ์‚ฝ์ž…, ์‹ค์‹œ๊ฐ„ ๋ฐฉํ–ฅ์„ฑ ์ถ”์ •์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ–ˆ๋‹ค. | | ํ•ต์‹ฌ ์•„์ด๋””์–ด โ€“ SWpC | 1๏ธโƒฃ ์˜ˆ์ธก ๋‹จ๊ณ„ : ์œˆ๋„์šฐ ๋‚ด ์ž…๋ ฅ ์‹ ํ˜ธ x ๋กœ๋ถ€ํ„ฐ ์ถœ๋ ฅ y ๋ฅผ LTI ์‹œ์Šคํ…œ(์ž„ํŽ„์Šค ์‘๋‹ต h )์„ ํ†ตํ•ด ์˜ˆ์ธก.<br>2๏ธโƒฃ ๊ฐ•๋„ ์ง€ํ‘œ

Network Quantitative Biology Model
No Image

Tomography by Design: An Algebraic Approach to Low-Rank Quantum States

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ ์–‘์ž ์ƒํƒœ ํ† ๋ชจ๊ทธ๋ž˜ํ”ผ(QST) ๋Š” ์‹œ์Šคํ…œ์˜ ๋ฐ€๋„ ํ–‰๋ ฌ ฯ(ํ—๋ฏธ์…˜, ์–‘์˜ ์ค€์ •๋ถ€ํ˜ธ, ํŠธ๋ ˆ์ด์Šค 1)๋ฅผ ์žฌ๊ตฌ์„ฑํ•˜๋Š” ํ•ต์‹ฌ ๊ธฐ์ˆ ์ด๋ฉฐ, ์ „ํ†ต์ ์ธ ์™„์ „ ์ธก์ • ๋ฐฉ์‹์€ ์ฐจ์› ์ˆ˜ D 2โฟ์— ๋Œ€ํ•ด O(Dยฒ)๊ฐœ์˜ ํŒŒ๋ผ๋ฏธํ„ฐ์™€ ์ธก์ • ์„ค์ •์„ ์š”๊ตฌํ•œ๋‹ค. ์ด๋Š” โ€œ์ฐจ์›์˜ ์ €์ฃผโ€๋ผ ๋ถˆ๋ฆฌ๋Š” ๋ฌธ์ œ๋ฅผ ์•ผ๊ธฐํ•œ๋‹ค. ์‹ค์ œ ๋ฌผ๋ฆฌ ์‹œ์Šคํ…œ์€ ์ €๊ณ„์ˆ˜, ์ €์—”ํŠธ๋กœํ”ผ, ๊ตญ์†Œ์„ฑ ๋“ฑ ๊ตฌ์กฐ์  ์ œ์•ฝ์„ ๊ฐ–๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ์•„, ์ด๋Ÿฌํ•œ ์ œ์•ฝ์„ ํ™œ์šฉํ•˜๋ฉด ์ธก์ •ยท๋ณต์› ๋น„์šฉ์„ ํฌ๊ฒŒ ๋‚ฎ์ถœ ์ˆ˜ ์žˆ๋‹ค. ๊ธฐ์กด ์—ฐ๊ตฌ๋Š” (i) ํ•ต์‹ฌโ€‘๋…ธ๋ฆ„์„ ์ด์šฉํ•œ ๋žญํฌ ์ตœ์†Œํ™” , (ii) ๊ณ ์ •โ€‘๋žญํฌ ํŒŒ๋ผ๋ฏธํ„ฐํ™”(Bu

Quantum Physics
No Image

Topological variations in General Relativity: a rigorous perspective

1. ์—ฐ๊ตฌ ๋™๊ธฐ์™€ ๋ฐฐ๊ฒฝ ์œ„์ƒ ๋™์—ญํ•™ : Wheeler์™€ Hawking์ด ์ œ์‹œํ•œ โ€œ์ŠคํŽ˜์ด์Šคโ€‘ํƒ€์ž„ ํผโ€ ์•„์ด๋””์–ด๋Š”, ์‹œ๊ณต๊ฐ„ ์ž์ฒด๊ฐ€ ์œ„์ƒ๊นŒ์ง€๋„ ๋™์ ์œผ๋กœ ๋ณ€ํ•  ์ˆ˜ ์žˆ๋‹ค๋Š” ๊ฐ€์„ค์„ ๋‹ด๊ณ  ์žˆ๋‹ค. ์ด๋Š” ์–‘์ž ์ค‘๋ ฅ์—์„œ โ€œ์ŠคํŽ˜์ด์Šคโ€‘ํƒ€์ž„ ํผโ€ ํ˜น์€ โ€œ์œ„์ƒ ์š”๋™โ€์ด๋ผ๋Š” ๊ฐœ๋…์œผ๋กœ ์žฌ์กฐ๋ช…๋œ๋‹ค. ์ˆ˜ํ•™์  ๊ณต๋ฐฑ : ๊ธฐ์กด ๋ฌผ๋ฆฌํ•™์  ๋…ผ์˜๋Š” ์œ„์ƒ ํ•จ์ˆ˜๋ฏธ๋ถ„(Topological functional derivative) ๋ฅผ ์ •์˜ํ•˜์ง€ ๋ชปํ•ด ๋ณ€๋ถ„ ์›๋ฆฌ ์ž์ฒด๊ฐ€ ํ˜•์‹์ ์œผ๋กœ๋งŒ ๋‚จ์•„ ์žˆ์—ˆ๋‹ค. ํŠนํžˆ โ€œ๋ฌดํ•œ์†Œ ์œ„์ƒ ๋ณ€๋™โ€์ด๋ผ๋Š” ๊ฐœ๋…์ด ๋ถ€์žฌํ–ˆ๊ธฐ ๋•Œ๋ฌธ์—, ๊ฒฝ๋กœ ์ ๋ถ„์—์„œ ์ •์  ์œ„์ƒ์ (sta

Mathematics
No Image

Towards a Humanized Social-Media Ecosystem: AI-Augmented HCI Design Patterns for Safety, Agency & Well-Being

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋ฌธ์ œ ์ •์˜ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์ค‘์‹ฌ ์„ค๊ณ„์˜ ํ•œ๊ณ„ : ๊ธฐ์กด ์†Œ์…œ ๋ฏธ๋””์–ด๋Š” โ€˜์ฐธ์—ฌ ๊ทน๋Œ€ํ™”โ€™๋ฅผ ๋ชฉํ‘œ๋กœ ์„ค๊ณ„๋ผ, ์‚ฌ์šฉ์ž์˜ ์ •์‹ ์ ยท์‚ฌํšŒ์  ๋น„์šฉ์„ ์™ธ๋ฉดํ•œ๋‹ค. ์‚ฌ์šฉ์ž ์ฃผ๋„ํ˜• ํ†ต์ œ ๋ถ€์žฌ : ํ”Œ๋žซํผ์ด ์ œ๊ณตํ•˜๋Š” ์„ค์ •์€ ์ผ๋ฐ˜์ ์œผ๋กœ โ€˜์ „์—ญ์ โ€™์ด๋ฉฐ, ์ƒํ™ฉ๋ณ„ ๋ฏธ์„ธ ์กฐ์ •์ด ์–ด๋ ต๋‹ค. 2. ํ•ต์‹ฌ ๊ฐœ๋… โ€“ Humanโ€‘Layer AI (HLโ€‘AI) | ํŠน์„ฑ | ์„ค๋ช… | | | | | ์‚ฌ์šฉ์ž ์†Œ์œ  | ๋ธŒ๋ผ์šฐ์ € ํ™•์žฅ ํ˜•ํƒœ๋กœ, ์‚ฌ์šฉ์ž๊ฐ€ ์ง์ ‘ ์„ค์น˜ยท์ œ์–ดํ•œ๋‹ค. | | ์„ค๋ช… ๊ฐ€๋Šฅ์„ฑ | ๊ฐ ์ œ์–ดยท์ถ”์ฒœ์— ๋Œ€ํ•œ ์ด์œ ๋ฅผ ์‹ค์‹œ๊ฐ„์œผ๋กœ ์ œ๊ณต, ํˆฌ๋ช…์„ฑ ํ™•๋ณด. | | ํ”Œ๋žซํผ ๋…๋ฆฝ์„ฑ

System
No Image

Towards a topological view of blood pressure regulation

1. ์—ฐ๊ตฌ์˜ ๋…์ฐฝ์„ฑ ๋ฐ ์˜์˜ ํ† ํด๋กœ์ง€ ์ค‘์‹ฌ ์ ‘๊ทผ : ๊ธฐ์กด ํ˜ˆ์•• ๋ชจ๋ธ์€ ํ˜ˆ๊ด€์„ โ€˜ํŠธ๋ฆฌ ๊ตฌ์กฐโ€™๋กœ ๊ฐ€์ •ํ•˜๊ณ  ๊ตญ์†Œ ์ €ํ•ญยท์ˆœ์‘์„ฑ์—๋งŒ ์ดˆ์ ์„ ๋งž์ท„๋‹ค. ๋ณธ ์—ฐ๊ตฌ๋Š” ํ์‡„ ๋ฃจํ”„ ๋ผ๋Š” ๋น„ํŠธ๋ฆฌ ๊ตฌ์กฐ๊ฐ€ ์ „์—ญ ์••๋ ฅ ๋ชจ๋“œ๋ฅผ ๋งŒ๋“ค ์ˆ˜ ์žˆ์Œ์„ ์ˆ˜ํ•™ยท์‹œ๋ฎฌ๋ ˆ์ด์…˜์ ์œผ๋กœ ์ฆ๋ช…ํ•จ์œผ๋กœ์จ, ํ˜ˆ์—ญํ•™ ์ด๋ก ์— ์ƒˆ๋กœ์šด ์ฐจ์›์„ ์ถ”๊ฐ€ํ•œ๋‹ค. ์ „์—ญโ€‘๊ตญ์†Œ ์ƒํ˜ธ์ž‘์šฉ : ๋ฃจํ”„ ๋‚ด๋ถ€์—์„œ๋Š” ๋ณ‘๋ ฌ ๊ฒฝ๋กœ๊ฐ€ ๋™์ ์œผ๋กœ ๊ฒฐํ•ฉ ๋˜์–ด, ํ•œ ์ง€์ ์˜ ์ €ํ•ญ ๋ณ€ํ™”๊ฐ€ ์ „์ฒด ๋ฃจํ”„์— ์˜ํ–ฅ์„ ๋ฏธ์นœ๋‹ค. ์ด๋Š” โ€œ์ „์—ญ ๋ชจ๋“œ โ‰  ๊ตญ์†Œ ์ €ํ•ญโ€์ด๋ผ๋Š” ๊ฐœ๋…์„ ์ •๋Ÿ‰ํ™”ํ•œ ์ฒซ ์‚ฌ๋ก€๋ผ ํ•  ์ˆ˜ ์žˆ๋‹ค. 2. ๋ชจ๋ธ๋งยท๋ฐฉ๋ฒ•๋ก  ํ‰๊ฐ€ | ํ•ญ๋ชฉ | ๊ฐ•์ 

Physics
No Image

Towards Universal Spatial Transcriptomics Super-Resolution: A Generalist Physically Consistent Flow Matching Framework

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ๊ณต๊ฐ„ ์ „์‚ฌ์ฒด ํ•ด์ƒ๋„์™€ ๋น„์šฉ์˜ ํŠธ๋ ˆ์ด๋“œ์˜คํ”„ ๋Š” ํ˜„์žฌ ST ์—ฐ๊ตฌ์˜ ๊ฐ€์žฅ ํฐ ๋ณ‘๋ชฉ์ด๋‹ค. ๊ธฐ์กด SR ๋ฐฉ๋ฒ•๋“ค์€ ์ด๋ฏธ์ง€ ์ดˆํ•ด์ƒ๋„์™€ ๋‹ฌ๋ฆฌ ์œ ์ „์ž ๋ฐœํ˜„๋Ÿ‰์ด๋ผ๋Š” ์งˆ๋Ÿ‰ ๋ณด์กด ๋ฒ•์น™ ์„ ๋ฌด์‹œํ•˜๊ณ  ์žˆ์–ด, ์‹ค์ œ ์ƒ๋ฌผํ•™์  ํ•ด์„์— ์œ„ํ—˜์„ ์ดˆ๋ž˜ํ•œ๋‹ค. ๋˜ํ•œ, ์ƒ๋ฌผํ•™์  ์ด์งˆ์„ฑ(์ข…, ์กฐ์ง, ๋ฐฐ์น˜ ์ฐจ์ด) ์ด ์‹ฌํ•ด OOD ์ƒํ™ฉ์—์„œ ๋ชจ๋ธ์ด ๊ธ‰๊ฒฉํžˆ ์„ฑ๋Šฅ ์ €ํ•˜๋ฅผ ๊ฒช๋Š”๋‹ค. 2. ํ•ต์‹ฌ ์•„์ด๋””์–ด ๋ฐ ๋ฐฉ๋ฒ•๋ก  | ๊ตฌ์„ฑ ์š”์†Œ | ์ฃผ์š” ์—ญํ•  | ๊ธฐ์ˆ ์  ํŠน์ง• | | | | | | Structureโ€‘Aware Semantic Alignment (SASA) | ์œ ์ „์ž

Framework Quantitative Biology
No Image

Tracking Time-Varying Multipath Channels forActive Sonar Applications

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ๋‹ค์ค‘๊ฒฝ๋กœ ๋ฐฐ๊ฒฝ์˜ ๋น„์ •์ƒ์„ฑ : ์–•์€ ํ•ด์—ญ์—์„œ๋Š” ํ‘œ๋ฉดยท๋ฐ”๋‹ฅ ๋ฐ˜์‚ฌ์— ์˜ํ•ด ๊ฐ•ํ•œ ์ž”ํ–ฅ์ด ๋ฐœ์ƒํ•˜๊ณ , ํŒŒ๋„์™€ ํ”Œ๋žซํผ ๋“œ๋ฆฌํ”„ํŠธ ๋•Œ๋ฌธ์— ํ†ต๊ณ„๊ฐ€ ์‹œ๊ฐ„์— ๋”ฐ๋ผ ๋ณ€ํ•œ๋‹ค. ์ด๋Š” CFAR ๊ธฐ๋ฐ˜ ๋ ˆ์ธ์ง€โ€‘๋„ํ”Œ๋Ÿฌ ์ฒ˜๋ฆฌ์˜ ๊ฐ€์ •(ํ†ต๊ณ„์  ์ •Stationarity)์„ ๊นจ๋œจ๋ ค ํƒ์ง€ ์„ฑ๋Šฅ์„ ์ €ํ•˜์‹œํ‚จ๋‹ค. ์›์‹œ ์ธก์ •๊ฐ’ ํ™œ์šฉ์˜ ์žฅ์  : ์œ„์ƒ ์ผ๊ด€์„ฑ์„ ์œ ์ง€ํ•˜๊ณ , ์ € SNR ์ƒํ™ฉ์—์„œ๋„ likelihoodโ€‘based ์ ‘๊ทผ์ด ๊ฐ€๋Šฅํ•˜๋‹ค๋Š” ์ ์—์„œ ๊ธฐ์กด ๋ณ€ํ™˜โ€‘ํ›„ ์ฒ˜๋ฆฌ๋ณด๋‹ค ์œ ๋ฆฌํ•˜๋‹ค. 2. ํ•ต์‹ฌ ์•„์ด๋””์–ด ๋ฐ ๋ชจ๋ธ๋ง | ๋‹จ๊ณ„ | ๋‚ด์šฉ | ์ฃผ์š” ์ˆ˜์‹/๊ฐœ๋… | | | |

Electrical Engineering and Systems Science
No Image

Training-Free Zero-Shot Anomaly Detection in 3D Brain MRI with 2D Foundation Models

1๏ธโƒฃ ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ์˜๋ฃŒ ์˜์ƒ ์ด์ƒ ํƒ์ง€ ๋Š” ์กฐ๊ธฐ ์ง„๋‹จยท์น˜๋ฃŒ ๊ณ„ํš์— ํ•ต์‹ฌ์ ์ด๋ฉฐ, ๊ธฐ์กด ๋น„์ง€๋„ ์žฌ๊ตฌ์„ฑ ๊ธฐ๋ฐ˜ ๋ฐฉ๋ฒ•(์˜คํ† ์ธ์ฝ”๋”, VAE, GAN, Diffusion ๋“ฑ)์€ ๋Œ€๊ทœ๋ชจ ์ •์ƒ ๋ฐ์ดํ„ฐ๊ฐ€ ํ•„์š”ํ•˜๊ณ , ์Šค์บ๋„ˆยทํ”„๋กœํ† ์ฝœ ๋ณ€ํ™”์— ์ทจ์•ฝํ•˜๋‹ค. Zeroโ€‘Shot Anomaly Detection (ZSAD) ๋Š” ์‚ฌ์ „ ํ•™์Šต๋œ ํŒŒ์šด๋ฐ์ด์…˜ ๋ชจ๋ธ์„ ๊ทธ๋Œ€๋กœ ํ™œ์šฉํ•ด ๋ผ๋ฒจ ์—†์ด๋„ ์ด์ƒ์„ ํƒ์ง€ํ•  ์ˆ˜ ์žˆ์–ด ๋น„์šฉยท์‹œ๊ฐ„ ์ ˆ๊ฐ ํšจ๊ณผ๊ฐ€ ํฌ๋‹ค. ํ•˜์ง€๋งŒ ํ˜„์žฌ๊นŒ์ง€๋Š” 2D ์ด๋ฏธ์ง€ ์—๋งŒ ์„ฑ๊ณต์ ์œผ๋กœ ์ ์šฉ๋ผ ์™”์œผ๋ฉฐ, 3D ์˜๋ฃŒ ์˜์ƒ์— ๊ทธ๋Œ€๋กœ ์ ์šฉํ•˜๋ฉด ๋ณผ๋ฅจ ๊ตฌ์กฐ ์†์‹ค ๊ณผ ๋ฉ”

Computer Science Computer Vision Detection Model
No Image

Transition from traveling fronts to diffusion-limited growth in expanding populations

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋ฌธ์ œ ์ œ๊ธฐ ์ „ํ†ต์  ๋ฐ˜์‘โ€‘ํ™•์‚ฐ ๋ชจ๋ธ (FKPP, ๋น„์„ ํ˜• ํ™•์‚ฐ ๋“ฑ)์€ ์ƒ์ˆ˜ ์†๋„ ์ „ํŒŒ ๋ฅผ ์˜ˆ์ธกํ•œ๋‹ค. ์ด๋Š” ์šด๋™์„ฑ ์ด ์žˆ๋Š” ๋ฏธ์ƒ๋ฌผ(์˜ˆ: ์ˆ˜์˜์„ฑ ๋ฐ•ํ…Œ๋ฆฌ์•„)์—์„œ๋Š” ์ž˜ ๋งž์ง€๋งŒ, ๋น„์šด๋™์„ฑ ์„ธํฌ ์ง‘๋‹จ ์—์„œ๋Š” ์ „ํŒŒ ์ „ํ˜•์ด ๋‹ค๋ฅด๋‹ค. ์‹คํ—˜์—์„œ๋Š” ๋ฐ˜์ง€๋ฆ„์ด tยนแŸยฒ ํ˜น์€ ๋ฉด์ ์ด t ๋กœ ์„ฑ์žฅํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋นˆ๋ฒˆํžˆ ๋ณด๊ณ ๋˜์—ˆ์œผ๋ฉฐ, ๊ธฐ์กด ๋ชจ๋ธ์€ ์ด๋ฅผ ์„ค๋ช…ํ•˜์ง€ ๋ชปํ•œ๋‹ค. 2. ๋ชจ๋ธ ์„ค๊ณ„ ํ•ต์‹ฌ ๊ฐ€์ • : ์ƒ๋ฌผ๋Ÿ‰ ์žฌ๋ถ„๋ฐฐ(ํ™•์‚ฐ) ์†๋„ Dโ‚š ๊ฐ€ ์„ฑ์žฅ๋ฅ (ฮณ n) ์— ๋น„๋ก€ํ•œ๋‹ค. ์ฆ‰, ์˜์–‘์†Œ๊ฐ€ ๋ถ€์กฑํ•˜๋ฉด ์„ธํฌ ์ด๋™๋„ ๊ฐ์†Œํ•œ๋‹ค๋Š” ์ ์„ ๋ฐ˜์˜ํ•œ๋‹ค. ๋น„์ฐจ์›ํ™” ํ›„ ํ•ต์‹ฌ ํŒŒ๋ผ

Physics
No Image

Translating Dietary Standards into Healthy Meals with Minimal Substitutions

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ๋น„๋งŒยท๋‹น๋‡จยท์‹ฌํ˜ˆ๊ด€์งˆํ™˜ ๋“ฑ ์ฃผ์š” ๋น„์ „์—ผ์„ฑ ์งˆํ™˜์˜ ์ฃผ์š” ์œ„ํ—˜ ์š”์ธ์ธ ์‹๋‹จ ์„ ๊ณผํ•™์  ๊ฐ€์ด๋“œ๋ผ์ธ(USDA)๊ณผ ์ผ์ƒ ์‹์‚ฌ ์‚ฌ์ด์— ๊ฒฉ์ฐจ๊ฐ€ ์กด์žฌํ•œ๋‹ค. ๊ธฐ์กด ๊ฐœ์ธํ™” ์‹๋‹จ ์ถ”์ฒœ ์‹œ์Šคํ…œ์€ ๋‹จ์ผ ๋ชฉํ‘œ (๋ง›, ์นผ๋กœ๋ฆฌ, ํŽธ๋ฆฌ์„ฑ) ์ตœ์ ํ™”์— ๋จธ๋ฌผ๋Ÿฌ ๊ฐ€์ด๋“œ๋ผ์ธ ์ค€์ˆ˜ ์—ฌ๋ถ€๋ฅผ ์ฒด๊ณ„์ ์œผ๋กœ ๊ฒ€์ฆํ•˜์ง€ ๋ชปํ•˜๊ณ , โ€˜์–ผ๋งˆ๋‚˜ ์ ๊ฒŒ ๋ฐ”๊ฟ”์•ผ ํ•˜๋Š”๊ฐ€โ€™ ๋ผ๋Š” ์‹ค์šฉ์  ์งˆ๋ฌธ์— ๋‹ตํ•˜์ง€ ๋ชปํ•œ๋‹ค. 2. ๋ฐ์ดํ„ฐ ๋ฐ ์ „์ฒ˜๋ฆฌ | ํ•ญ๋ชฉ | ๋‚ด์šฉ | | | | | ๋ฐ์ดํ„ฐ ์ถœ์ฒ˜ | NHANES WWEIA (2013โ€‘2020, 6 wave) | | ์ด ์‹์‚ฌ ์ˆ˜ | 135 49

Computer Science Artificial Intelligence
No Image

Tunable Ferroelectric Acoustic Resonators in Monolithic Thin-Film Barium Titanate

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ๋ฌด์„  ํ†ต์‹ ์˜ ํŠธ๋ Œ๋“œ : 5G/6G ๋ฐ IoT ํ™•๋Œ€๋กœ ๋‹ค์ค‘ ๋Œ€์—ญยท๊ณ ๋Œ€์—ญํญ ์š”๊ตฌ๊ฐ€ ๊ธ‰์ฆํ•˜๊ณ  ์žˆ๋‹ค. ๊ธฐ์กด ์ „์ž๊ธฐ(EM) ํ•„ํ„ฐ๋Š” ํฌ๊ธฐยท์†์‹คยท๋น„์šฉ ์ธก๋ฉด์—์„œ ํ•œ๊ณ„๊ฐ€ ์žˆ๋‹ค. ์ดˆ์ŒํŒŒ ํ•„ํ„ฐ์˜ ์žฅ์  : ์†Œํ˜•ยท์ €์‚ฝ์ž…์†์‹คยท๋‹ค์ค‘ ๊ณต์ง„๊ธฐ ๊ตฌํ˜„์ด ์šฉ์ดํ•ด ๋ชจ๋ฐ”์ผยท์นฉ ์ˆ˜์ค€์—์„œ์˜ ์ง‘์ ์— ์ ํ•ฉํ•˜๋‹ค. ๊ฐ€๋ณ€ํ˜• ํ•„์š”์„ฑ : ํ•˜๋‚˜์˜ ํ•„ํ„ฐ๋กœ ์—ฌ๋Ÿฌ ๋Œ€์—ญ์„ ์ปค๋ฒ„ํ•˜๋ฉด ๋ถ€ํ’ˆ ์ˆ˜ยท๋ณด๋“œ ๋ฉด์ ยท์ „๋ ฅ ์†Œ๋ชจ๋ฅผ ํฌ๊ฒŒ ์ค„์ผ ์ˆ˜ ์žˆ๋‹ค. ๊ธฐ์กด ๊ฐ€๋ณ€ํ˜• ๊ธฐ์ˆ (์ƒ๋ณ€ํ™” ๋ฌผ์งˆ, ํŽ˜๋กญ์ž์„ฑ, MEMS varactor ๋“ฑ)์€ ๋ณต์žกํ•œ ๊ณต์ •ยท์ „์••ยท์‹ ๋ขฐ์„ฑ ๋ฌธ์ œ๋ฅผ ์•ˆ๊ณ  ์žˆ๋‹ค. 2. ํ•ต์‹ฌ ์•„์ด๋””

Electrical Engineering and Systems Science
No Image

Tunable microwave frequency synthesis with optically-derived spectral purity

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๊ธฐ์กด ํ•œ๊ณ„ ์ „ํ†ต์  ๊ฐ€๋ณ€ ๋งˆ์ดํฌ๋กœํŒŒ ์†Œ์Šค (YIG, DRO ๋“ฑ)๋Š” ๋„“์€ ํŠœ๋‹ ๋ ˆ์ธ์ง€์™€ ์ €์œ„์ƒ ์žก์Œ ์‚ฌ์ด์—์„œ ํƒ€ํ˜‘์ด ํ•„์š”ํ–ˆ๋‹ค. ๊ณ ํ’ˆ์งˆ ๊ณ ์ • ์ฃผํŒŒ์ˆ˜ ๋ ˆ์กฐ๋„ค์ดํ„ฐ(์‚ฌํŒŒ์ด์–ด ์บ๋น„ํ‹ฐ ๋“ฑ)๋Š” ์ˆœ์ˆ˜์„ฑ์„ ์ œ๊ณตํ•˜์ง€๋งŒ ๊ฐ€๋ณ€์„ฑ์ด ๊ฑฐ์˜ ์—†๋‹ค. ๊ด‘ํ•™ ์ฃผํŒŒ์ˆ˜ ๋ถ„ํ• (OFD) ์€ ๊ด‘ํ•™ ๋ ˆ์ด์ €์˜ ์ดˆ๊ณ โ€‘Q๋ฅผ ๋งˆ์ดํฌ๋กœํŒŒ๋กœ ๋‚˜๋ˆ„์–ด ์ „์ดํ•จ์œผ๋กœ์จ ์ตœ๊ณ ์˜ ์ˆœ์ˆ˜์„ฑ์„ ์–ป์ง€๋งŒ, ๊ณ ์ • ์ฃผํŒŒ์ˆ˜ ์ถœ๋ ฅ์ด ๊ธฐ๋ณธ์ด๋ฉฐ, ์‹คํ—˜์‹ค ์ˆ˜์ค€์˜ ๋ณต์žกํ•œ ์žฅ๋น„(์ดˆ์•ˆ์ • ๋ ˆ์ด์ €, ์…€ํ”„โ€‘๋ ˆํผ๋Ÿฐ์‹ฑ comb)๊ฐ€ ํ•„์š”ํ–ˆ๋‹ค. eOFD ๋Š” ๋‘ ๊ด‘ํ•™ ํ†ค์„ ์ด์šฉํ•ด ์ „๊ธฐโ€‘๊ด‘ ๋ณ€์กฐ๊ธฐ๋กœ ๋งˆ์ดํฌ๋กœํŒŒ๋ฅผ ๊ด‘ํ•™ ์ฐจ์ด๊นŒ

Physics
No Image

Two-way Clustering Robust Variance Estimator in Quantile Regression Models

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ ๋‹ค์ค‘ ๊ตฐ์ง‘ ๋ฐ์ดํ„ฐ : ๋…ธ๋™ยท๊ธฐ์—…, ์ˆ˜์ถœยท๋ชฉ์ ์ง€ ๋“ฑ ๋‘ ์ฐจ์›(ํ–‰ยท์—ด)์—์„œ ๋™์‹œ์— ์ƒ๊ด€๊ด€๊ณ„๊ฐ€ ์กด์žฌํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋นˆ๋ฒˆํžˆ ๋‚˜ํƒ€๋‚œ๋‹ค. ๊ธฐ์กด์˜ ์ผ๋ฐฉํ–ฅ(clusterโ€‘robust) ๋ฐฉ๋ฒ•์€ ์ด๋Ÿฌํ•œ ๋ณตํ•ฉ ์˜์กด์„ฑ์„ ์ถฉ๋ถ„ํžˆ ๋ฐ˜์˜ํ•˜์ง€ ๋ชปํ•œ๋‹ค. ๋ถ„์œ„์ˆ˜ ํšŒ๊ท€์˜ ํ•„์š”์„ฑ : ํ‰๊ท  ํšŒ๊ท€์™€ ๋‹ฌ๋ฆฌ QR์€ ์กฐ๊ฑด๋ถ„ํฌ์˜ ๊ผฌ๋ฆฌยท์ด์งˆ์„ฑ์„ ์ง์ ‘ ๋ถ„์„ํ•  ์ˆ˜ ์žˆ์–ด ๊ธˆ์œตยท๊ฒฝ์ œ ๋ถ„์•ผ์—์„œ ํ•ต์‹ฌ ๋„๊ตฌ์ด๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ QR์˜ ์ ์ˆ˜ ํ•จ์ˆ˜๋Š” ๋น„์—ฐ์†์ ์ด๋ฉฐ, ๊ตฐ์ง‘ ๊ตฌ์กฐ์™€ ๊ฒฐํ•ฉ๋  ๋•Œ ์ด๋ก ์  ๋‚œ์ œ๊ฐ€ ํฌ๊ฒŒ ์ฆ๊ฐ€ํ•œ๋‹ค. 2. ํ•ต์‹ฌ ์ด๋ก ์  ํ”„๋ ˆ์ž„์›Œํฌ | ์š”์†Œ | ๋‚ด์šฉ | ์˜์˜ | | |

Economics Model
No Image

UltraLIF: Fully Differentiable Spiking Neural Networks via Ultradiscretization and Max-Plus Algebra

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ SNN์˜ ์žฅ์  : ์ด๋ฒคํŠธโ€‘๋“œ๋ฆฌ๋ธ ์—ฐ์‚ฐ์œผ๋กœ ์ €์ „๋ ฅ ํ•˜๋“œ์›จ์–ด(Loihi ๋“ฑ)์™€ ์ž์—ฐ์Šค๋Ÿฌ์šด ์‹œ๊ฐ„ ์ •๋ณด ์ฒ˜๋ฆฌ์— ์œ ๋ฆฌ. ํ•™์Šต ๋‚œ์ œ : ์ŠคํŒŒ์ดํฌ ๋ฐœ์ƒ์„ ๋‚˜ํƒ€๋‚ด๋Š” Heaviside ํ•จ์ˆ˜๋Š” ๋ฏธ๋ถ„์ด ๋ถˆ๊ฐ€๋Šฅํ•ด, ๊ธฐ์กด์—๋Š” surrogate gradient (๋Œ€์ฒด ๊ทธ๋ž˜๋””์–ธํŠธ) ๋ฐฉ์‹์— ์˜์กด. ์ด๋Š” ์ „๋ฐฉโ€‘ํ›„๋ฐฉ ๋ถˆ์ผ์น˜ (forwardโ€‘backward mismatch)์™€ ์ด๋ก ์  ์ˆ˜๋ ด ๋ณด์žฅ์ด ๋ถ€์กฑํ•˜๋‹ค๋Š” ํ•œ๊ณ„๊ฐ€ ์žˆ๋‹ค. 2. ํ•ต์‹ฌ ์•„์ด๋””์–ด โ€“ ์šธํŠธ๋ผ๋””์Šคํฌ๋ฆฌํƒ€์ด์ œ์ด์…˜ ์šธํŠธ๋ผ๋””์Šคํฌ๋ฆฌํƒ€์ด์ œ์ด์…˜ ์€ ์—ฐ์† ๋ฏธ๋ถ„ ๋ฐฉ์ •์‹์„ maxโ€‘plus ๋ฐ˜๋Œ€์ฒด (tr

Computer Science Network Machine Learning
No Image

Uncertainty-Aware Neural Multivariate Geostatistics

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ๋‹ค๋ณ€๋Ÿ‰ ๊ณต๊ฐ„ ๋ฐ์ดํ„ฐ ๋Š” ํ™˜๊ฒฝยท์ƒํƒœยท์›๊ฒฉํƒ์‚ฌ ๋ถ„์•ผ์—์„œ ์ ์  ๋” ํ”ํ•ด์ง€๊ณ  ์žˆ๋‹ค. ๋ณ€์ˆ˜ ๊ฐ„ ์ƒ๊ด€๊ด€๊ณ„์™€ ๊ณต๊ฐ„ ์ž๊ธฐ์ƒ๊ด€์„ ๋™์‹œ์— ๋ชจ๋ธ๋งํ•ด์•ผ ํ•˜์ง€๋งŒ, ์ „ํ†ต์ ์ธ ๋‹ค๋ณ€๋Ÿ‰ Gaussian Process (GP) ๋‚˜ Linear Model of Coregionalization (LMC) ์€ 1) ์ •์ƒ์„ฑ ๊ฐ€์ • (๊ณต๊ฐ„์  ์ƒ๊ด€์ด ๊ฑฐ๋ฆฌ๋งŒ ์˜์กด) 2) ๊ณ„์‚ฐ ๋ณต์žก๋„ (O(nยณ)ยทO(Jยฒnยณ) ๋“ฑ) ์— ํ•œ๊ณ„๊ฐ€ ์žˆ๋‹ค. ์ตœ๊ทผ ๋น„์ •์ƒ์„ฑ ์„ ๋ฐ˜์˜ํ•˜๊ธฐ ์œ„ํ•œ ๋‹ค์–‘ํ•œ ์ปค๋ฒ„๋ฆฌ์–ธ์Šค ๊ตฌ์กฐ(๊ณต๊ฐ„์  Matรฉrn ํŒŒ๋ผ๋ฏธํ„ฐ ๋ณ€๋™, ์ปค๋„ ์ปจ๋ณผ๋ฃจ์…˜, ์ž ์žฌ ์ฐจ์› ์ ‘

Statistics
No Image

Uni-Flow: a unified autoregressive-diffusion model for complex multiscale flows

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋ฌธ์ œ ์ •์˜ ๋‹ค์ค‘์Šค์ผ€์ผ ํ๋ฆ„ ์€ ํฐ ์Šค์ผ€์ผ(์ „์ฒด ํ๋ฆ„ ๊ตฌ์กฐ)๊ณผ ์ž‘์€ ์Šค์ผ€์ผ(์™€๋ฅ˜ยท๊ฒฝ๊ณ„์ธต) ์‚ฌ์ด์˜ ์ƒํ˜ธ์ž‘์šฉ์ด ํ•ต์‹ฌ์ด๋ฉฐ, ์ „ํ†ต์ ์ธ CFD(LES, DNS)๋Š” ํ•ด์ƒ๋„ยท์‹œ๊ฐ„ ์Šคํ… ์š”๊ตฌ๋Ÿ‰์ด ํญ๋ฐœ์ ์œผ๋กœ ์ฆ๊ฐ€ํ•œ๋‹ค. ๊ธฐ์กด ๋ฐ์ดํ„ฐโ€‘๊ธฐ๋ฐ˜ ์ ‘๊ทผ : AR/์ˆœํ™˜ ์‹ ๊ฒฝ๋ง ์€ ์‹œ๊ฐ„ ์˜์กด์„ฑ์„ ์ž˜ ํฌ์ฐฉํ•˜์ง€๋งŒ ์žฅ๊ธฐ ๋กค์•„์›ƒ์—์„œ ๋ฐœ์‚ฐํ•˜๊ฑฐ๋‚˜ ๋ฌผ๋ฆฌ์  ์ผ๊ด€์„ฑ์„ ์žƒ๋Š”๋‹ค. ์‹ ๊ฒฝ ์—ฐ์‚ฐ์ž(FOU, DeepONet ๋“ฑ) ๋Š” ๋งคํ•‘ ์ž์ฒด๋Š” ํšจ์œจ์ ์ด์ง€๋งŒ ๊ณ ํ•ด์ƒ๋„ ์žฌ๊ตฌ์„ฑ์— ํ•œ๊ณ„๊ฐ€ ์žˆ๋‹ค. ํ™•์‚ฐ ๋ชจ๋ธ ์€ ์ด๋ฏธ์ง€ยท์‹œ๋ฎฌ๋ ˆ์ด์…˜ ๋ณต์›์— ๊ฐ•์ ์ด ์žˆ์œผ๋‚˜ ์‹œ๊ฐ„ ํ๋ฆ„์„ ์ง์ ‘ ๋‹ค๋ฃจ์ง€ ๋ชปํ•œ๋‹ค

Physics Model
No Image

Unified Eigenvalue-Eigenspace Criteria for Functional Properties of Linear Systems and the Generalized Separation Principle

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ์ „ํ†ต์  ์ ‘๊ทผ์˜ ํ•œ๊ณ„ : ๊ณ ์ „์ ์ธ ์ œ์–ด๊ฐ€๋Šฅ์„ฑยท๊ด€์ธก๊ฐ€๋Šฅ์„ฑ์€ ์ „์ฒด ์ƒํƒœ (x(t))์— ์ดˆ์ ์„ ๋งž์ถ”์–ด, ๋Œ€๊ทœ๋ชจ ์‹œ์Šคํ…œ์—์„œ๋Š” ์ฐจ์› ํญ๋ฐœ๊ณผ ๊ณ„์‚ฐ ๋ณต์žก๋„๊ฐ€ ๊ธ‰์ฆํ•œ๋‹ค. ๊ธฐ๋Šฅ์  ์š”๊ตฌ : ์ „๋ ฅ ์‹œ์Šคํ…œ์˜ ์ด ์ „๋ ฅ, ๊ตํ†ต ๋„คํŠธ์›Œํฌ์˜ ์ „์ฒด ํ๋ฆ„, ์ƒํƒœ๊ณ„ ๋ชจ๋ธ์˜ ์ด ๊ฐœ์ฒด์ˆ˜ ๋“ฑ ์„ ํ˜• ์กฐํ•ฉ ๋งŒ์„ ์ œ์–ดยท๊ด€์ธกํ•˜๋ฉด ์ถฉ๋ถ„ํ•œ ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค. ์ด๋Š” functional controllability/observability ๋ผ๋Š” ์ƒˆ๋กœ์šด ๊ฐœ๋…์„ ์ •๋‹นํ™”ํ•œ๋‹ค. 2. ํ•ต์‹ฌ ๊ธฐ์—ฌ | ๋ฒˆํ˜ธ | ๋‚ด์šฉ | ๊ธฐ์กด ์—ฐ๊ตฌ์™€ ์ฐจ๋ณ„์  | | | | | | 1 | ํ†ตํ•ฉ

System Electrical Engineering and Systems Science
No Image

Uniform error bounds for quantized dynamical models

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ ์‹œ์Šคํ…œ ์‹๋ณ„ ๊ณผ ํ•™์Šต ์ด๋ก  ์„ ์—ฐ๊ฒฐํ•˜๋Š” ์ „ํ†ต์  ์ ‘๊ทผ์€ ๋Œ€๋ถ€๋ถ„ i.i.d. ๋ฐ์ดํ„ฐ ๊ฐ€์ •์— ๊ธฐ๋ฐ˜ํ•œ๋‹ค. ์‹ค์ œ ์ œ์–ดยท์ž„๋ฒ ๋””๋“œ ์‹œ์Šคํ…œ์—์„œ๋Š” ์‹œ๊ฐ„ ์˜์กด์„ฑ(ฮฒโ€‘mixing) ๊ณผ ์–‘์žํ™”(์ €๋น„ํŠธ ๊ตฌํ˜„) ๊ฐ€ ํ•„์ˆ˜์ ์ด๋ฉฐ, ์ด ๋‘ ์š”์†Œ๋ฅผ ๋™์‹œ์— ๊ณ ๋ คํ•œ ๋น„๋Œ€์นญ์  ์ด๋ก ์€ ๋ถ€์กฑํ–ˆ๋‹ค. ํŠนํžˆ ํ•˜์ด๋ธŒ๋ฆฌ๋“œ ์‹œ์Šคํ…œ (์ˆจ์€ ์ „์ดยท๋‹ค์ค‘ ์„œ๋ธŒ์‹œ์Šคํ…œ)์—์„œ๋Š” ์•Œ๊ณ ๋ฆฌ์ฆ˜โ€‘ํŠนํ™” ๋ถ„์„์ด ์–ด๋ ค์›Œ ์•Œ๊ณ ๋ฆฌ์ฆ˜โ€‘๋ถˆ๋ณ€(uniform) ๋ณด์žฅ์ด ์š”๊ตฌ๋œ๋‹ค. 2. ์ฃผ์š” ๊ธฐ์—ฌ | ๋ฒˆํ˜ธ | ๋‚ด์šฉ | ์˜์˜ | | | | | | โ‘  | ํ†ตํ•ฉ ์˜ค๋ฅ˜ ํ•œ๊ณ„ (Uniform error bo

Computer Science Machine Learning Model
No Image

UniTAF: A Modular Framework for Joint Text-to-Speech and Audio-to-Face Modeling

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ | ๊ธฐ์กด ํŒŒ์ดํ”„๋ผ์ธ | ๋ฌธ์ œ์  | | | | | LLM โ†’ TTS โ†’ A2F | ๊ฐ์ •ยทํ”„๋กœ์†Œ๋””์™€ ๊ฐ™์€ ๊ณ ์ˆ˜์ค€ ์ •๋ณด๊ฐ€ ํ…์ŠคํŠธยท์˜ค๋””์˜ค์—๋งŒ ๊ตญํ•œ๋ผ ๋‘ ๋ฒˆ์”ฉ ์ถ”๋ก  โ†’ ์ •๋ณด ์†์‹คยท์—ฐ์‚ฐ ์ค‘๋ณต | | Endโ€‘toโ€‘End ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ ๋ชจ๋ธ | ์ถ”๋ก  ๋น„์šฉยท์ œ์–ด์„ฑ ๋ถ€์กฑ, ์‹ค์„œ๋น„์Šค ์ ์šฉ ์–ด๋ ค์›€ | TTS ์ตœ์‹  ๋ชจ๋ธ(IndexTTS2)์€ ์ค‘๊ฐ„ ํ‘œํ˜„ (duration, pitch, acoustic token ๋“ฑ)์„ ๋ช…์‹œ์ ์œผ๋กœ ์ƒ์„ฑํ•œ๋‹ค. ์ด๋Š” LLM์ด ๋‚ดํฌํ•œ ์˜๋„์™€ ๊ฐ์ •์„ ๊ตฌ์กฐํ™”ํ•œ ํ˜•ํƒœ๋กœ ๋ณผ ์ˆ˜ ์žˆ๋‹ค. ๋”ฐ๋ผ์„œ ์ด ์ •๋ณด๋ฅผ A2F์— ์ง์ ‘

Computer Science Framework Sound Model
No Image

Universal Approximation Theorems for Dynamical Systems with Infinite-Time Horizon Guarantees

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋ฌธ์ œ ์ •์˜ ๊ธฐ์กด ํ•œ๊ณ„ : RNNยทNeural ODE์— ๋Œ€ํ•œ ๋ณดํŽธ ๊ทผ์‚ฌ ์ •๋ฆฌ๋Š” ์ฃผ๋กœ ์œ ํ•œ ์‹œ๊ฐ„ ํ˜น์€ ์ „์—ญ์  ์•ˆ์ •์„ฑ( fadingโ€‘memory ) ์„ ์ „์ œ๋กœ ํ•œ๋‹ค. ์ด๋Š” ๋‹ค์ค‘ ์•ˆ์ •์„ฑ(์—ฌ๋Ÿฌ ๋Œ๊ฐœ)์ด๋‚˜ ๋ฆฌ๋ฐ‹ ์‚ฌ์ดํด์„ ํฌํ•จํ•˜๋Š” ์‹ค์ œ ๋‡Œยท์ œ์–ด ์‹œ์Šคํ…œ์„ ์„ค๋ช…ํ•˜์ง€ ๋ชปํ•œ๋‹ค. ์„ธ ๊ฐ€์ง€ ์˜ค๋ฅ˜ ์œ ํ˜• Bโ€‘type (Basin mismatch) : ์ดˆ๊ธฐ์กฐ๊ฑด์ด ๋ถ„๋ฆฌ๋ฉด(separatrix) ๊ทผ์ฒ˜์— ์žˆ์„ ๋•Œ ์ž‘์€ ๋ฒกํ„ฐ์žฅ ์˜ค์ฐจ๊ฐ€ ๋‹ค๋ฅธ ๋Œ๊ฐœ๋กœ ์ด๋Œ์–ด ํฐ ๊ถค์  ์ฐจ์ด๋ฅผ ๋งŒ๋“ ๋‹ค. Pโ€‘type (Phase drift) : ๋ฆฌ๋ฐ‹ ์‚ฌ์ดํด์˜ ์ฃผ๊ธฐ ์ฐจ์ด๋กœ ์ธํ•ด

System Mathematics
No Image

Universal Beta Splatting

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ Gaussian Splatting ์€ ์ตœ๊ทผ ๋ช‡ ๋…„๊ฐ„ ๊ณ ํ’ˆ์งˆ ์‹ค์‹œ๊ฐ„ ์‹ ๊ฒฝ ๋ Œ๋”๋ง์—์„œ ํฐ ์ฃผ๋ชฉ์„ ๋ฐ›์•˜์ง€๋งŒ, ๊ณ ์ •๋œ ํ˜•ํƒœ (๋“ฑ๋ฐฉ์„ฑ, ์ •๊ทœ๋ถ„ํฌ) ๋•Œ๋ฌธ์— ๋ณต์žกํ•œ ๊ด‘ํ•™ ํ˜„์ƒ(์˜ˆ: ๊ฐ•ํ•œ ๋ฐ˜์‚ฌ, ๋น„๋“ฑ๋ฐฉ์„ฑ ์žฌ์งˆ, ๋™์  ๋ณ€ํ™”)์„ ํ‘œํ˜„ํ•˜๋Š” ๋ฐ ํ•œ๊ณ„๊ฐ€ ์žˆ์—ˆ๋‹ค. ๊ธฐ์กด ํ•ด๊ฒฐ์ฑ…์€ ๋ณด์กฐ ๋„คํŠธ์›Œํฌ (์˜ˆ: ๋ށ์Šค, ๋…ธ๋ฉ€, ๋ผ์ดํŠธ ๋„คํŠธ์›Œํฌ)๋‚˜ ๋‹ค์ค‘ ์ฑ„๋„ ์ธ์ฝ”๋”ฉ ์„ ๋„์ž…ํ•ด ๋ณด์™„ํ–ˆ์ง€๋งŒ, ์ด๋Š” ํŒŒ๋ผ๋ฏธํ„ฐ ์ˆ˜์™€ ์—ฐ์‚ฐ ๋น„์šฉ์„ ๊ธ‰๊ฒฉํžˆ ๋Š˜๋ ธ๋‹ค. 2. ํ•ต์‹ฌ ์•„์ด๋””์–ด โ€“ ๋ฒ ํƒ€ ์ปค๋„ Beta ๋ถ„ํฌ ๋Š” (

No Image

Universal priors: solving empirical Bayes via Bayesian inference and pretraining

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ ๊ฒฝํ—˜์  ๋ฒ ์ด์ฆˆ(Empirical Bayes, EB) ๋Š” ์‚ฌ์ „๋ถ„ํฌ๋ฅผ ์ง์ ‘ ์ถ”์ •ํ•˜์ง€ ์•Š๊ณ , ๊ด€์ธก ๋ฐ์ดํ„ฐ ์ „์ฒด๋ฅผ ์ด์šฉํ•ด ๊ฐ ํŒŒ๋ผ๋ฏธํ„ฐ๋ฅผ ์ถ”์ •ํ•จ์œผ๋กœ์จ ์ „ํ†ต์ ์ธ MLE๋ณด๋‹ค ๋‚ฎ์€ ์œ„ํ—˜์„ ๋‹ฌ์„ฑํ•œ๋‹ค. ๊ธฐ์กด ๋ฐฉ๋ฒ•(Robbins estimator, NPMLE, gโ€‘modeling ๋“ฑ)์€ ๊ฐ ๋ฌธ์ œ๋งˆ๋‹ค ๋ณ„๋„ ํ•™์Šต ์ด ํ•„์š”ํ•˜๊ฑฐ๋‚˜ ๊ณ„์‚ฐ ๋น„์šฉ์ด ํฌ๊ฒŒ ๋“ ๋‹ค. ์ตœ๊ทผ TabPFN ยท Transformer ๊ธฐ๋ฐ˜ ์‚ฌ์ „ํ•™์Šต ์ ‘๊ทผ๋ฒ•์ด โ€œํ•œ ๋ฒˆ ํ•™์Šตํ•˜๋ฉด ๋‹ค์–‘ํ•œ ๋ฐ์ดํ„ฐ์…‹์— ๋ฐ”๋กœ ์ ์šฉ ๊ฐ€๋Šฅโ€ํ•˜๋‹ค๋Š” ์ ์—์„œ ์ฃผ๋ชฉ๋ฐ›๊ณ  ์žˆ๋‹ค. 2. ํ•ต์‹ฌ ๊ธฐ์—ฌ | ๋ฒˆํ˜ธ | ๋‚ด์šฉ

Statistics Machine Learning
No Image

Validating Interpretability in siRNA Efficacy Prediction: A Perturbation-Based, Dataset-Aware Protocol

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ siRNA ์น˜๋ฃŒ์ œ ๊ฐ€ FDA ์Šน์ธ์„ ๋ฐ›์œผ๋ฉฐ ์ž„์ƒ์  ์ค‘์š”์„ฑ์ด ์ปค์กŒ๊ณ , ์ปดํ“จํ„ฐ ๊ธฐ๋ฐ˜ ํšจ๋Šฅ ์˜ˆ์ธก ์ด ํ›„๋ณด ์„ ๋ณ„ ๋น„์šฉ์„ ํฌ๊ฒŒ ์ ˆ๊ฐํ•œ๋‹ค. ์ตœ์‹  ๋”ฅ๋Ÿฌ๋‹ ๋ชจ๋ธ์€ ๋†’์€ ์˜ˆ์ธก ์ •ํ™•๋„๋ฅผ ๋ณด์ด์ง€๋งŒ, ์„ค๋ช…(์‚ด๋ฆฌ์–ธ์‹œ) ํ™œ์šฉ ์ด ์‹ค์ œ ์‹คํ—˜ ์„ค๊ณ„์— ์ง์ ‘ ์—ฐ๊ฒฐ๋˜๋Š” ์ƒํ™ฉ์—์„œ ์„ค๋ช…์˜ ์‹ ๋ขฐ์„ฑ ์ด ๊ฒ€์ฆ๋˜์ง€ ์•Š์œผ๋ฉด ์˜คํžˆ๋ ค ๋น„์šฉ์„ ์ฆ๊ฐ€์‹œํ‚ฌ ์œ„ํ—˜์ด ์žˆ๋‹ค. 2. ํ•ต์‹ฌ ์ œ์•ˆ โ€“ โ€œpreโ€‘synthesis gateโ€ | ๋‹จ๊ณ„ | ๋‚ด์šฉ | ํ•ต์‹ฌ ํฌ์ธํŠธ | | | | | | (i) Saliency ๊ณ„์‚ฐ | Gradient magnitude์„ nucleot

Data Quantitative Biology
No Image

VaultGemma: A Differentially Private Gemma Model

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ํ”„๋ผ์ด๋ฒ„์‹œ ์œ„ํ—˜ : ๋Œ€ํ˜• ์–ธ์–ด ๋ชจ๋ธ(LLM)์€ ๋ฐฉ๋Œ€ํ•œ ์›น ํ…์ŠคํŠธ๋ฅผ ํ•™์Šตํ•˜๋ฉด์„œ ๊ฐœ์ธ ์‹๋ณ„ ์ •๋ณด(PII)๋ฅผ ๋‚ดํฌํ•  ์œ„ํ—˜์ด ์žˆ๋‹ค. ์ตœ๊ทผ ๋ฐ์ดํ„ฐ ์œ ์ถœยท์—ญ์ถ”์  ์‚ฌ๋ก€๊ฐ€ ์ฆ๊ฐ€ํ•˜๋ฉด์„œ, ๋ชจ๋ธ ํ•™์Šต ๋‹จ๊ณ„์—์„œ ํ”„๋ผ์ด๋ฒ„์‹œ ๋ณดํ˜ธ๊ฐ€ ํ•„์ˆ˜์ ์œผ๋กœ ์š”๊ตฌ๋œ๋‹ค. ์ฐจ๋“ฑ ํ”„๋ผ์ด๋ฒ„์‹œ(DP) : DP๋Š” ํ†ต๊ณ„์  ๋ณด์žฅ์„ ์ œ๊ณตํ•ด ๊ฐœ๋ณ„ ๋ฐ์ดํ„ฐ ๋ ˆ์ฝ”๋“œ๊ฐ€ ๋ชจ๋ธ ์ถœ๋ ฅ์— ๋ฏธ์น˜๋Š” ์˜ํ–ฅ์„ ์ œํ•œํ•œ๋‹ค. ํŠนํžˆ, ฮตโ€‘DP(ํ”„๋ผ์ด๋ฒ„์‹œ ์˜ˆ์‚ฐ)์™€ ฮด(์‹คํŒจ ํ™•๋ฅ ) ํŒŒ๋ผ๋ฏธํ„ฐ๋ฅผ ํ†ตํ•ด ๋ณด์•ˆ ์ˆ˜์ค€์„ ์ •๋Ÿ‰ํ™”ํ•œ๋‹ค. 2. VaultGemma 1B์˜ ํ•ต์‹ฌ ๊ธฐ์—ฌ | ํ•ญ๋ชฉ | ๋‚ด์šฉ | | | |

Model
No Image

Vintage Code, Modern Judges: Meta-Validation in Low Data Regimes

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ๋ ˆ๊ฑฐ์‹œ ์‹œ์Šคํ…œ ํ˜„๋Œ€ํ™”๋Š” ๊ธฐ์—…์˜ ๋””์ง€ํ„ธ ์ „ํ™˜์—์„œ ํ•ต์‹ฌ ๊ณผ์ œ์ด์ง€๋งŒ, COBOLยทPL/IยทREXX ๋“ฑ ์˜ค๋ž˜๋œ ์–ธ์–ด์— ๋Œ€ํ•œ ์ „๋ฌธ๊ฐ€๊ฐ€ ๊ธ‰๊ฐํ•˜๊ณ  ์žˆ๋‹ค. ์ธ๊ฐ„ ํ‰๊ฐ€ ๋ฐ์ดํ„ฐ๊ฐ€ ๋ถ€์กฑํ•œ ์ƒํ™ฉ์—์„œ LLM์„ โ€œ์‹ฌํŒโ€์œผ๋กœ ํ™œ์šฉํ•˜๋ ค๋Š” ์‹œ๋„๋Š” ์ž์—ฐ์Šค๋Ÿฝ์ง€๋งŒ, ๊ฒ€์ฆ๋˜์ง€ ์•Š์€ ์‹ฌํŒ์„ ๊ทธ๋Œ€๋กœ ์‹ ๋ขฐํ•˜๋ฉด ํ‰๊ฐ€ ์ˆœํ™˜ ์˜ค๋ฅ˜(evaluation loop) ๊ฐ€ ๋ฐœ์ƒํ•œ๋‹ค๋Š” ์œ„ํ—˜์„ฑ์„ ์ •ํ™•ํžˆ ์งš๊ณ  ์žˆ๋‹ค. 2. ํ•ต์‹ฌ ๊ธฐ์—ฌ SparseAlign ํ”„๋ ˆ์ž„์›Œํฌ : pairwiseโ€‘confidence : ๋‘ ์ƒ˜ํ”Œ ๊ฐ„ ์ƒ๋Œ€์  ์ˆœ์œ„๊ฐ€ ์–ผ๋งˆ๋‚˜ ํ™•์‹  ์žˆ๊ฒŒ ํŒ๋‹จ๋˜๋Š”์ง€๋ฅผ ์ •๋Ÿ‰

Data
No Image

Virtualized 3D Gaussians: Flexible Cluster-based Level-of-Detail System for Real-Time Rendering of Composed Scenes

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ 3D Gaussian Splatting ์€ ์ตœ๊ทผ ๊ณ ํ’ˆ์งˆ ์‹ ๊ฒฝ ๋ Œ๋”๋ง ๋ถ„์•ผ์—์„œ ์ฃผ๋ชฉ๋ฐ›๋Š” ๊ธฐ์ˆ ์ด๋ฉฐ, ํŠนํžˆ ๋ช…์‹œ์  ์ด๊ณ  ์ด์‚ฐ์ ์ธ ํ”„๋ฆฌ๋ฏธํ‹ฐ๋ธŒ ๋•๋ถ„์— ์”ฌ ํ•ฉ์„ฑ์ด ์šฉ์ดํ•˜๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ์Šค์ผ€์ผ ์ด ์ปค์งˆ์ˆ˜๋ก(์ˆ˜์–ต ๊ฐœ ๊ฐ€์šฐ์‹œ์•ˆ) ๋ฉ”๋ชจ๋ฆฌยท์—ฐ์‚ฐ ๋ถ€ํ•˜๊ฐ€ ๊ธ‰์ฆํ•ด ์‹ค์‹œ๊ฐ„ ๋ Œ๋”๋ง์ด ๋ถˆ๊ฐ€๋Šฅํ•ด์ง„๋‹ค. ๊ธฐ์กด LOD(๋ ˆ๋ฒจโ€‘์˜ค๋ธŒโ€‘๋””ํ…Œ์ผ) ๊ธฐ๋ฒ•์€ ์ฃผ๋กœ ๋ฉ”์‹œ ๊ธฐ๋ฐ˜ ์ด๊ฑฐ๋‚˜ ๋ณผ๋ฅจ ๊ธฐ๋ฐ˜ ์ด๋ฉฐ, ๊ฐ€์šฐ์‹œ์•ˆ ํ”„๋ฆฌ๋ฏธํ‹ฐ๋ธŒ์— ์ง์ ‘ ์ ์šฉํ•˜๊ธฐ์—” ๊ตฌ์กฐ์  ์ฐจ์ด๊ฐ€ ํฌ๋‹ค. 2. ํ•ต์‹ฌ ์•„์ด๋””์–ด Nanite ์™€ ์œ ์‚ฌํ•˜๊ฒŒ ํด๋Ÿฌ์Šคํ„ฐ ๊ธฐ๋ฐ˜ ๊ฐ€์ƒํ™” ๋ฅผ ๋„์ž…ํ•ด, ๊ฐ€์šฐ์‹œ์•ˆ์„ ๊ณ„์ธต์  ํด๋Ÿฌ์Šคํ„ฐ

System
No Image

Waveform Design for ISAC System: A Consensus ADMM Approach

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ISAC ์€ 5Gยท6G ์‹œ๋Œ€์— ์ŠคํŽ™ํŠธ๋Ÿผ ํšจ์œจ, ์ง€์—ฐ ๊ฐ์†Œ, ํ•˜๋“œ์›จ์–ด ๊ณต์œ  ๋“ฑ ๋‹ค์ค‘ ์ด์ ์„ ์ œ๊ณตํ•œ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ํ†ต์‹  (๊ณ ์† ๋ฐ์ดํ„ฐ ์ „์†ก)๊ณผ ๋ ˆ์ด๋” (๊ณ ๊ฐ๋„ ๋ชฉํ‘œ ํƒ์ง€)๋ผ๋Š” ์„œ๋กœ ๋‹ค๋ฅธ ๋ชฉํ‘œ๋ฅผ ๋™์‹œ์— ๋งŒ์กฑ์‹œํ‚ค๋Š” ํŒŒํ˜• ์„ค๊ณ„๋Š” ๋ณธ์งˆ์ ์œผ๋กœ ๋‹ค๋ชฉ์  ๋น„๋ณผ๋ก ์ตœ์ ํ™” ๋ฌธ์ œ๋ฅผ ๋งŒ๋“ ๋‹ค. ๊ธฐ์กด ์—ฐ๊ตฌ๋“ค์€ ๋‹จ์ผ ๋ชฉํ‘œ ์ค‘์‹ฌ (ํ†ต์‹ โ€‘์ค‘์‹ฌ ํ˜น์€ ๊ฐ์ง€โ€‘์ค‘์‹ฌ) ์„ค๊ณ„, ํ˜น์€ ๋ธ”๋ก ์ขŒํ‘œ ํ•˜๊ฐ•(Blockโ€‘Coordinate Descent) , ๊ฐ€์ค‘ ์ตœ์†Œ ํ‰๊ท  ์ œ๊ณฑ์˜ค์ฐจ(WMMSE) ๋“ฑ์œผ๋กœ ์ ‘๊ทผํ–ˆ์ง€๋งŒ, CM ์ œ์•ฝ ๊ณผ ํŒŒํ˜• ์œ ์‚ฌ์„ฑ ์„ ๋™์‹œ์— ๊ณ ๋ คํ•œ ํ†ตํ•ฉ

System Electrical Engineering and Systems Science
No Image

What Do Neurons Listen To? A Neuron-level Dissection of a General-purpose Audio Model

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ ๋ฒ”์šฉ ์˜ค๋””์˜ค SSL ๋ชจ๋ธ ์€ ๋‹ค์–‘ํ•œ ๋‹ค์šด์ŠคํŠธ๋ฆผ ์ž‘์—…์—์„œ ๋›ฐ์–ด๋‚œ ์„ฑ๋Šฅ์„ ๋ณด์ด์ง€๋งŒ, ๋‚ด๋ถ€ ํ‘œํ˜„์ด ์–ด๋–ป๊ฒŒ ์ผ๋ฐ˜ํ™”๋ฅผ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๋Š”์ง€ ๋Š” ์•„์ง ๋ฏธ์ง€์ˆ˜์ด๋‹ค. ์ตœ๊ทผ ๊ธฐ๊ณ„ํ•ด์„(mechanistic interpretability) ๋ถ„์•ผ๊ฐ€ ๋Œ€ํ˜• ์–ธ์–ด ๋ชจ๋ธ(LLM)์—์„œ ๋‰ด๋Ÿฐโ€‘๋‹จ์œ„ ๋ถ„์„์„ ํ†ตํ•ด ์˜๋ฏธยท์†์„ฑโ€‘ํŠน์ด ๋‰ด๋Ÿฐ์„ ์ฐพ์•„๋‚ด๋Š” ๋ฐฉ๋ฒ•์„ ์ œ์‹œํ–ˆ์œผ๋ฉฐ, ์ด๋ฅผ ์˜ค๋””์˜ค ์˜์—ญ์— ์ ์šฉํ•˜๊ณ ์ž ํ•˜๋Š” ์‹œ๋„๊ฐ€ ๋ณธ ๋…ผ๋ฌธ์˜ ํ•ต์‹ฌ ๋™๊ธฐ๋‹ค. 2. ํ•ต์‹ฌ ๋ฐฉ๋ฒ•๋ก  | ๋‹จ๊ณ„ | ๋‚ด์šฉ | ๋ชฉ์  | | | | | | A. ๋‰ด๋Ÿฐ ์ •์˜ | Transformer Encode

Audio Processing Electrical Engineering and Systems Science Model
No Image

When to repeat a biomarker test? Decomposing sources of variation from conditionally repeated measurements

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ์ž„๊ณ„๊ฐ’ ๊ธฐ๋ฐ˜ ์žฌ๊ฒ€์‚ฌ์˜ ์œ„ํ—˜ : ํ˜ˆ์•ก ๊ธฐ์ฆ ์ „ Hb๊ฐ€ ๊ธฐ์ค€์น˜ ์ดํ•˜์ผ ๊ฒฝ์šฐ ์žฌ์ธก์ •ํ•˜์ง€๋งŒ, ์žฌ์ธก์ • ์ž์ฒด๊ฐ€ โ€œ์กฐ๊ฑด๋ถ€โ€์ด๊ธฐ ๋•Œ๋ฌธ์— ๊ด€์ธก๋œ ๋ฐ์ดํ„ฐ๋Š” ์„ ํƒ ํŽธํ–ฅ์„ ๋‚ดํฌํ•œ๋‹ค. ์ด๋Š” ์ž„์ƒ์‹œํ—˜์—์„œ ํ”ํžˆ ๋…ผ์˜๋˜๋Š” โ€œsequential testing biasโ€์™€ ๋™์ผํ•œ ๋ฉ”์ปค๋‹ˆ์ฆ˜์ด๋‹ค. ์ธก์ •์˜ค์ฐจ์™€ ์ธ๊ตฌ ๋ณ€๋™์„ฑ ๊ตฌ๋ถ„์˜ ์ค‘์š”์„ฑ : Hb์™€ ๊ฐ™์€ ์—ฐ์†ํ˜• ๋ฐ”์ด์˜ค๋งˆ์ปค๋Š” ์ง„๋‹จยท์น˜๋ฃŒ ๊ฒฐ์ •์— ์ž„๊ณ„๊ฐ’์„ ์ ์šฉํ•œ๋‹ค. ์ธก์ •์˜ค์ฐจ๊ฐ€ ํด ๊ฒฝ์šฐ falseโ€‘positive(๋ถˆํ•„์š”ํ•œ ๊ธฐ์ฆ ๊ฑฐ์ ˆ)์™€ falseโ€‘negative(๋นˆํ˜ˆ ๊ธฐ์ฆ์ž ํ—ˆ์šฉ) ์œ„ํ—˜์ด ๋™์‹œ์— ์ฆ๊ฐ€ํ•œ

Statistics Applications
No Image

Where to Search: Measure the Prior-Structured Search Space of LLM Agents

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ Iterative paradigm (์ƒ์„ฑโ€‘ํ•„ํ„ฐโ€‘์ •์ œ)๋Š” LLM์ด ์ดˆ๊ธฐ ์•„์ด๋””์–ด๋ฅผ ์ƒ์„ฑํ•˜๊ณ , ์ด๋ฅผ ๊ฒ€์ฆยท์ˆ˜์ •ํ•˜๋Š” ๊ณผ์ •์„ ๋ฐ˜๋ณตํ•จ์œผ๋กœ์จ ๋ณต์žกํ•œ ๋ฌธ์ œ ํ•ด๊ฒฐ์— ๊ฐ•์ ์„ ๋ณด์ธ๋‹ค. ๊ธฐ์กด ์—ฐ๊ตฌ๋Š” โ€˜๋ฌด์—‡์„ ์ƒ์„ฑํ•˜๋А๋ƒโ€™ ์— ์ดˆ์ ์„ ๋งž์ท„์ง€๋งŒ, โ€˜์–ด๋””์„œ ํƒ์ƒ‰ํ•˜๋А๋ƒโ€™ โ€”์ฆ‰ ํƒ์ƒ‰ ๊ณต๊ฐ„์˜ ๊ตฌ์กฐ์™€ ์ œ์•ฝโ€”์— ๋Œ€ํ•œ ์ •๋Ÿ‰์  ๋…ผ์˜๋Š” ๋ถ€์กฑํ–ˆ๋‹ค. ๋„๋ฉ”์ธ ์‚ฌ์ „(์˜ˆ: ๋ฌผ๋ฆฌ ๋ฒ•์น™, ํ™”ํ•™ ๋ฐ˜์‘ ๊ทœ์น™ ๋“ฑ)์„ ํ˜•์‹ํ™” ํ•˜๊ณ  ์ด๋ฅผ ํƒ์ƒ‰ ๊ณต๊ฐ„์— ๋ฐ˜์˜ํ•˜๋Š” ๊ฒƒ์ด, ํƒ์ƒ‰ ํšจ์œจ๊ณผ ์•ˆ์ „์„ฑ์„ ๋™์‹œ์— ๋ณด์žฅํ•˜๋Š” ํ•ต์‹ฌ ๊ณผ์ œ๋กœ ๋– ์˜ค๋ฅธ๋‹ค. 2. ํ•ต์‹ฌ ์ด๋ก  ๊ตฌ์กฐ | ๊ตฌ์„ฑ ์š”์†Œ | ์„ค

No Image

Who Evaluates AI's Social Impacts? Mapping Coverage and Gaps in First and Third Party Evaluations

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ๊ธฐ์ดˆ ๋ชจ๋ธ(FM)์˜ ํ™•๋Œ€ : GPTโ€‘4, PaLM ๋“ฑ ๋Œ€๊ทœ๋ชจ ์‚ฌ์ „ํ•™์Šต ๋ชจ๋ธ์ด ๋‹ค์–‘ํ•œ ๊ณ ์œ„ํ—˜ ์„œ๋น„์Šค(์˜๋ฃŒ, ๋ฒ•๋ฅ , ๊ธˆ์œต ๋“ฑ)์— ์ ์šฉ๋˜๋ฉด์„œ ์‚ฌํšŒ์  ์œ„ํ—˜์ด ๊ธ‰์ฆํ•˜๊ณ  ์žˆ๋‹ค. ๊ฑฐ๋ฒ„๋„Œ์Šค ์˜์กด๋„ ์ฆ๊ฐ€ : EU AI Act, ๋ฏธ๊ตญ AI Bill of Rights ๋“ฑ ๊ทœ์ œ ์ดˆ์•ˆ์ด โ€œํ‰๊ฐ€(evaluation)โ€๋ฅผ ํ•ต์‹ฌ ์š”๊ฑด์œผ๋กœ ๋ช…์‹œํ•จ์— ๋”ฐ๋ผ ํ‰๊ฐ€ ์ž๋ฃŒ์˜ ์งˆยท์–‘์ด ์ •์ฑ… ๊ฒฐ์ •์— ์ง์ ‘์ ์ธ ์˜ํ–ฅ์„ ๋ฏธ์นœ๋‹ค. ํ‰๊ฐ€ ๊ฒฉ์ฐจ ์ธ์‹ : ๊ธฐ์กด ์—ฐ๊ตฌ๋Š” ๋ชจ๋ธ ์„ฑ๋Šฅ(accuracy, robustness) ์ค‘์‹ฌ์˜ โ€œcapability evaluati

No Image

Whodunnit? The case of midge swarms

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•ต์‹ฌ ์งˆ๋ฌธ ๊ด€์ธก๋œ ์—ญ์„ค : ํŒŒ๋ฆฌ ๋–ผ๋Š” ํŽธ๊ทน(polarization) ๊ฐ’์ด ๊ฑฐ์˜ 0์— ๊ฐ€๊น๊ณ , ๊ฐœ์ฒด ๊ฐ„ ๊ฒฐํ•ฉ์ด ์•ฝํ•จ์—๋„ ๋ถˆ๊ตฌํ•˜๊ณ  ์ž์—ฐ์—์„œ๋Š” ์ƒ๊ด€๊ธธ์ด๊ฐ€ ๊ตฐ์ง‘ ์ „์ฒด์— ๊ฑธ์ณ ์Šค์ผ€์ผโ€‘ํ”„๋ฆฌํ•˜๊ฒŒ ๋Š˜์–ด๋‚œ๋‹ค. ์‹คํ—˜์‹ค vs ํ˜„์žฅ : ์‹คํ—˜์‹ค์—์„œ๋Š” ์™ธ๋ถ€ ๊ต๋ž€์ด ์—†์„ ๋•Œ ์ƒ๊ด€๊ธธ์ด๊ฐ€ ์งง๊ณ , ๊ต๋ž€์„ ๊ฐ€ํ•˜๋ฉด ์Šค์ผ€์ผโ€‘ํ”„๋ฆฌ๊ฐ€ ํšŒ๋ณต๋œ๋‹ค. ์ด๋Š” ์™ธ๋ถ€ ํ™˜๊ฒฝ์ด ๊ตฐ์ง‘ ์—ญํ•™์— ๊ฒฐ์ •์ ์ธ ์—ญํ• ์„ ํ•จ์„ ์‹œ์‚ฌํ•œ๋‹ค. ํ•ต์‹ฌ ๋ชฉํ‘œ : โ€œ๋ˆ„๊ฐ€ ๋ฒ”์ธ์ธ๊ฐ€?โ€๋ผ๋Š” ๋ฉ”ํƒ€ํฌ๋ฅผ ๋นŒ๋ ค, ๋„ค ๊ฐ€์ง€ ๊ธฐ์กด ๋ชจ๋ธ์ด ๊ด€์ธก๋œ ํ˜„์ƒ์„ ์–ผ๋งˆ๋‚˜ ์ž˜ ์„ค๋ช…ํ•˜๋Š”์ง€ ํ‰๊ฐ€ํ•˜๊ณ , ๊ฐ€์žฅ ์œ ๋ ฅํ•œ ํ›„๋ณด๋ฅผ ์ฐพ์•„๋‚ด๋Š”

Condensed Matter
No Image

Willchain: Decentralized, Privacy-Preserving, Self-Executing, Digital Wills

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ๋””์ง€ํ„ธ ์ž์‚ฐ ์ฆ๊ฐ€ : ์•”ํ˜ธํ™”ํ, NFT, ๋””์ง€ํ„ธ ๊ณ„์•ฝ ๋“ฑ ๊ฐœ์ธ์ด ๋ณด์œ ํ•˜๋Š” ๋””์ง€ํ„ธ ์ž์‚ฐ์ด ๊ธ‰์ฆํ•˜๋ฉด์„œ ์ „ํ†ต์ ์ธ ์œ ์–ธยท์ƒ์† ์ œ๋„๋Š” ๊ธฐ์ˆ ์  ํ•œ๊ณ„์— ์ง๋ฉดํ•˜๊ณ  ์žˆ๋‹ค. ํ”„๋ผ์ด๋ฒ„์‹œ์™€ ์‹ ๋ขฐ ๋ฌธ์ œ : ๊ธฐ์กด ์ค‘์•™ํ™”๋œ ์„œ๋น„์Šค๋Š” ๊ฐœ์ธ ์ •๋ณด ์œ ์ถœ ์œ„ํ—˜๊ณผ ์‹ ๋ขฐ์„ฑ ๋ถ€์กฑ์„ ์ดˆ๋ž˜ํ•œ๋‹ค. ํƒˆ์ค‘์•™ํ™”์™€ ์•”ํ˜ธํ•™์  ํ”„๋ผ์ด๋ฒ„์‹œ ๋ณดํ˜ธ๋Š” ์ด๋Ÿฌํ•œ ๋ฌธ์ œ๋ฅผ ๊ทผ๋ณธ์ ์œผ๋กœ ํ•ด๊ฒฐํ•  ์ˆ˜ ์žˆ๋‹ค. 2. ํ•ต์‹ฌ ๊ธฐ์ˆ ์  ๊ธฐ์—ฌ | ๊ตฌ๋ถ„ | ๋‚ด์šฉ | ์˜์˜ | | | | | | ๋ ˆ์ด์–ดโ€‘1 ํ”„๋กœํ† ์ฝœ | ์ž์ฒด ํผ๋ฏธ์…˜๋ฆฌ์Šค ์ฒด์ธ ์„ค๊ณ„, ๊ฒ€์ฆ์ธ ๋„คํŠธ์›Œํฌ์™€ ์ธํ„ฐ์ฒด์ธ ๋ฆด๋ ˆ์ด์–ด ํ™œ์šฉ | ๊ธฐ์กด L

No Image

ZeroSyl: Simple Zero-Resource Syllable Tokenization for Spoken Language Modeling

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ Pure Speech LM ์€ ํ…์ŠคํŠธ ์ž์›์ด ๋ถ€์กฑํ•œ ์–ธ์–ด์— ๋Œ€ํ•œ NLP ์ ์šฉ์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜์ง€๋งŒ, ํ˜„์žฌ๊นŒ์ง€ syntactic ์„ฑ๋Šฅ์ด ์ •์ฒด๋ผ ์žˆ๋‹ค(2023๋…„ ์ดํ›„ ์ •์ฒด). ๊ธฐ์กด SSLโ€‘๊ธฐ๋ฐ˜ ํ† ํฐ(์˜ˆ: w2vโ€‘BERT XL)์€ ํ”„๋ ˆ์ž„โ€‘๋ ˆ๋ฒจ (โ‰ˆ100 Hz) ํ† ํฐ์„ ์ƒ์„ฑํ•ด ์‹œํ€€์Šค ๊ธธ์ด๊ฐ€ ํ…์ŠคํŠธ ๋Œ€๋น„ 10~20๋ฐฐ ๊ธธ์–ด์ง โ†’ ์žฅ๊ฑฐ๋ฆฌ ์˜์กด์„ฑ ํ•™์Šต ๋น„์šฉ ๊ธ‰์ฆ . ์Œ์ ˆโ€‘๋‹จ์œ„ ํ† ํฐ์€ ์‹œํ€€์Šค ์••์ถ• ๊ณผ ์Œ์„ฑโ€‘์–ธ์–ด ๊ตฌ์กฐ ์‚ฌ์ด์˜ ์ข‹์€ ์ ˆ์ถฉ์ ์œผ๋กœ ์ œ์‹œ๋์ง€๋งŒ, Sylber, SyllableLM ๋“ฑ์€ ๋‹ค๋‹จ๊ณ„ ํŒŒ์ธํŠœ๋‹ + ํŠน์ˆ˜ ๋ชฉ์  ์†์‹ค

Model Computer Science NLP
No Image

A Class of algebras admitting infinitely many norm topologies

** ๋ณธ ๋…ผ๋ฌธ์—์„œ๋Š” ๋ณต์†Œ์ˆ˜ ์ฒด ์œ„์˜ ์—ฐ๊ด€ ๋Œ€์ˆ˜ (mathcal{A})์— ๋Œ€ํ•ด, ๊ณฑ์œผ๋กœ ์ƒ์„ฑ๋œ ๋ถ€๋ถ„๋Œ€์ˆ˜ (mathcal{A}^{2}= operatorname{span}{ab : a,binmathcal{A}})์˜ ์—ฌ์ฐจ์›(codimension)์ด ๋ฌดํ•œํ•  ๊ฒฝ์šฐ์™€ โ€œ๋ถˆ์—ฐ์† ์ œ๊ณฑ ์†Œ๋ฉธ ์„ฑ์งˆ(DSAP, Discontinuous Square Annihilation Property)โ€์„ ๊ฐ–๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋™์น˜์ž„์„ ์ฆ๋ช…ํ•œ๋‹ค. ํŠนํžˆ (mathcal{A}^{2})๊ฐ€ (mathcal{A}) ์•ˆ์—์„œ ๋ฌดํ•œํžˆ ํฐ ์—ฌ์ฐจ์›์„ ๊ฐ€์งˆ ๋•Œ, (mathcal{A})๋Š” ์„œ๋กœ

Mathematics
No Image

A Fully Discrete Nonnegativity-Preserving FEM for a Stochastic Heat Equation

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ SPDE์—์„œ ๋น„์Œ์ˆ˜์„ฑ : ์ž…์ž ๋ฐ€๋„ยท๋†๋„ ๋“ฑ ๋ฌผ๋ฆฌ๋Ÿ‰์€ ๋ณธ์งˆ์ ์œผ๋กœ ๋น„์Œ์ˆ˜์ด์–ด์•ผ ํ•˜๋ฏ€๋กœ, ์ˆ˜์น˜ ์Šคํ‚ค๋งˆ๊ฐ€ ์ด ๊ตฌ์กฐ์  ํŠน์„ฑ์„ ์œ ์ง€ํ•˜๋Š” ๊ฒƒ์ด ์žฅ๊ธฐ ์‹œ๋ฎฌ๋ ˆ์ด์…˜์˜ ๋ฌผ๋ฆฌ์  ์ผ๊ด€์„ฑ๊ณผ ์•ˆ์ •์„ฑ์— ํ•„์ˆ˜์ ์ด๋‹ค. ๊ธฐ์กด ์—ฐ๊ตฌ ํ•œ๊ณ„ : ์ „ํ†ต์ ์œผ๋กœ ๋น„์Œ์ˆ˜ ๋ณด์กด ์Šคํ‚ค๋งˆ๋Š” ์œ ํ•œ ์ฐจ๋ถ„๋ฒ•์— ๊ตญํ•œ๋ผ ์žˆ์—ˆ์œผ๋ฉฐ, ์ฐจ์›ยท์ง€์˜ค๋ฉ”ํŠธ๋ฆฌ ๋ณต์žก๋„์— ๋Œ€ํ•œ ํ™•์žฅ์„ฑ์ด ๋ถ€์กฑํ–ˆ๋‹ค. 2. ์ฃผ์š” ๊ธฐ์—ฌ | ๋ฒˆํ˜ธ | ๋‚ด์šฉ | ํ•ต์‹ฌ ์ •๋ฆฌ | | | | | | 1 | ๋น„์„ ํ˜• ์ผ๋ฐ˜ํ™” : ๊ธฐ์กด ์„ ํ˜•( (f(u) lambda u) ) ์Šคํ‚ค๋งˆ๋ฅผ Lipschitz ์—ฐ์† ๋น„์„ ํ˜• (f

Mathematics
No Image

A higher order pressure-stabilized virtual element formulation for the Stokes-Poisson-Boltzmann equations

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ์ „๊ธฐ์œ ์ฒด ํ˜„์ƒ (electrokinetics)์€ ๋‚˜๋…ธํฌ์–ดยท๋งˆ์ดํฌ๋กœํ”Œ๋ฃจ์ด๋”• ๋””๋ฐ”์ด์Šค ์„ค๊ณ„์— ํ•ต์‹ฌ์ด๋ฉฐ, Stokesโ€‘Poissonโ€‘Boltzmann(SPB) ์—ฐ๋™ ๋ฐฉ์ •์‹ ์„ ํ’€์–ด์•ผ ํ•œ๋‹ค. ์ „ํ†ต์ ์ธ Taylorโ€‘Hood FEM ์€ (i) ์„œ๋กœ ๋‹ค๋ฅธ ์ฐจ์ˆ˜์˜ ํ˜ผํ•ฉ ๊ณต๊ฐ„ ํ•„์š”, (ii) hangingโ€‘node ์ฒ˜๋ฆฌ ์–ด๋ ค์›€, (iii) ๋ณต์žกํ•œ ๊ฒฝ๊ณ„ยท๋‹คํ˜• ๊ฒฉ์ž์— ๋Œ€ํ•œ ์ œํ•œ ๋“ฑ ์‹ค์šฉ์ ์ธ ์ œ์•ฝ์ด ์žˆ๋‹ค. ๊ฐ€์ƒ์š”์†Œ๋ฒ•(VEM) ์€ ๋‹คํ˜• ๊ฒฉ์ž ์™€ hangingโ€‘node ์„ ์ž์—ฐ์Šค๋Ÿฝ๊ฒŒ ๋‹ค๋ฃฐ ์ˆ˜ ์žˆ์œผ๋ฉด์„œ, ๋™์ฐจ ์ฐจ์ˆ˜ ๊ทผ์‚ฌ๋„ ๊ฐ€๋Šฅํ•˜๋‹ค๋Š”

Mathematics
No Image

A New Lower Bound for the Diagonal Poset Ramsey Numbers

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋ฌธ์ œ ์„ค์ • ํฌ์…‹ ๋ผ๋จธ์ˆ˜ ๋Š” ๊ณ ์ „์ ์ธ ๊ทธ๋ž˜ํ”„ ๋ผ๋จธ์ˆ˜์˜ ๋ถ€๋ถ„์ง‘ํ•ฉ ๊ฒฉ์ž(ํฌ์…‹) ๋ฒ„์ „์ด๋ฉฐ, ํ•˜์ดํผํ๋ธŒ (Q N) ์—์„œ์˜ ์ƒ‰์น ์„ ํ†ตํ•ด ์ •์˜๋œ๋‹ค. ๋Œ€๊ฐ์„  ๋ผ๋จธ์ˆ˜ (R(Q n,Q n)) ์€ ๊ฐ€์žฅ ๊ธฐ๋ณธ์ ์ธ ๊ฒฝ์šฐ์ด์ง€๋งŒ, ํ˜„์žฌ๊นŒ์ง€๋Š” ์ƒํ•œ๊ณผ ํ•˜ํ•œ ์‚ฌ์ด์— ํฐ ๊ฒฉ์ฐจ๊ฐ€ ์กด์žฌํ•œ๋‹ค. ๊ธฐ์กด ํ•˜ํ•œ (2n) ์€ โ€œ๋ ˆ๋ฒจ ๊ตฌ๋ถ„โ€ ์ƒ‰์น (ํฌ๊ธฐ๊ฐ€ ( le n 1) ์ธ ๋ ˆ๋ฒจ์„ ํŒŒ๋ž‘, ๊ทธ ์ด์ƒ์„ ๋นจ๊ฐ•) ์—์„œ ๋ฐ”๋กœ ์–ป์–ด์ง€๋Š” trivial bound์ด๋ฉฐ, ์ด๋ฅผ ๊ฐœ์„ ํ•˜๋ ค๋ฉด ๋ ˆ๋ฒจ ๊ฐ„์˜ ๋ฏธ์„ธ ์กฐ์ •์ด ํ•„์š”ํ•˜๋‹ค. 2. ์ฃผ์š” ์•„์ด๋””์–ด์™€ ๊ธฐ๋ฒ• 1. ๋‹ค์ธต(la

Mathematics
No Image

A remark on staircase laminates in restricted sets

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ ๋ณผ๋ก ์ ๋ถ„(Convex Integration) ์€ ๋น„์„ ํ˜• ํŽธ๋ฏธ๋ถ„๋ฐฉ์ •์‹์˜ โ€œ๋น„์ •๊ทœ ํ•ดโ€๋ฅผ ๊ตฌ์ถ•ํ•˜๋Š” ๊ฐ•๋ ฅํ•œ ๋„๊ตฌ์ด๋ฉฐ, ์ตœ๊ทผ ๋ผ๋ฏธ๋„ค์ดํŠธ(laminate) ๊ตฌ์กฐ๋ฅผ ์ด์šฉํ•œ ์ •๋Ÿ‰์  ๋ฒ„์ „์ด ํ™œ๋ฐœํžˆ ์—ฐ๊ตฌ๋˜๊ณ  ์žˆ๋‹ค. Kleinerโ€‘Mรผllerโ€‘Szรฉkelyhidiโ€‘Xie(2023)๋Š” Lแต–โ€‘reducibility ๊ฐœ๋…์„ ๋„์ž…ํ•ด, โ€œ์ง‘ํ•ฉ K ๋กœ์˜ ๋ฏธ์„ธํ•œ ๋ผ๋ฏธ๋„ค์ดํŠธ ๋ณ€ํ™˜์ด ๊ฐ€๋Šฅํ•˜๋ฉด Kโ€‘๋‚ด ์ •ํ™•ํ•œ ํ•ด๋ฅผ ๋งŒ๋“ค ์ˆ˜ ์žˆ๋‹คโ€๋Š” ์ผ๋ฐ˜์ ์ธ ์ •๋ฆฌ๋ฅผ ์ œ์‹œํ–ˆ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ๊ทธ๋“ค์˜ ์˜ˆ์‹œ๋“ค์€ U โ„^{dร—m} ํ˜น์€ U๊ฐ€ K์™€ ํฌ๊ฒŒ ๊ฒน์น˜์ง€ ์•Š๋Š” ๊ฒฝ์šฐ ์—

Mathematics
No Image

A Systematic Evaluation of Sample-Level Tokenization Strategies for MEG Foundation Models

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ํ•„์š”์„ฑ ์‹ ๊ฒฝ์˜์ƒ ๋Œ€ํ˜• ๋ชจ๋ธ(LNM) : EEG/MEG์™€ ๊ฐ™์€ ๊ณ ํ•ด์ƒ๋„ ์ „๊ธฐ์ƒ๋ฆฌํ•™ ๋ฐ์ดํ„ฐ๋Š” ๋Œ€๊ทœ๋ชจ ๋น„๋ผ๋ฒจ ๋ฐ์ดํ„ฐ๊ฐ€ ํ’๋ถ€ํ•˜์ง€๋งŒ, ์—ฐ์†์ ์ธ ์‹œ๊ณ„์—ด์„ ์ง์ ‘ ๋‹ค๋ฃจ๊ธฐ์—” ๊ณ„์‚ฐ๋Ÿ‰์ด ๋ฐฉ๋Œ€ํ•ฉ๋‹ˆ๋‹ค. ๋”ฐ๋ผ์„œ ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ์—์„œ ์ฐจ์šฉํ•œ ํ† ํฌ๋‚˜์ด์ œ์ด์…˜์ด ํ•„์ˆ˜์ ์ž…๋‹ˆ๋‹ค. ํ† ํฌ๋‚˜์ด์ œ์ด์…˜์˜ ์„ค๊ณ„ ์˜๋ฌธ : ๊ธฐ์กด ํ† ํฌ๋‚˜์ด์ €๋Š” ๊ธˆ์œตยท์Œ์„ฑยท์‹œ๊ณ„์—ด ๋“ฑ ๋น„์ƒ๋ฌผํ•™์  ๋ฐ์ดํ„ฐ์— ์ตœ์ ํ™”๋œ ๊ฒƒ์ด ๋Œ€๋ถ€๋ถ„์ด๋ฉฐ, ๋‡Œ ์‹ ํ˜ธ์˜ ํŠน์ˆ˜ํ•œ ์ฃผํŒŒ์ˆ˜ยท์ง„ํญ ๋ถ„ํฌ๋ฅผ ์ถฉ๋ถ„ํžˆ ๋ฐ˜์˜ํ•˜์ง€ ๋ชปํ•  ๊ฐ€๋Šฅ์„ฑ์ด ์žˆ์Šต๋‹ˆ๋‹ค. 2. ์ฃผ์š” ๊ธฐ์—ฌ 1. ํ•™์Šต ๊ฐ€๋Šฅํ•œ ์ƒ˜ํ”Œโ€‘๋ ˆ๋ฒจ ํ† ํฌ๋‚˜์ด์ € ์ œ์•ˆ ์ž๋™์ธ์ฝ”๋”(AE) ๊ธฐ๋ฐ˜

Computer Science System Machine Learning Model
No Image

A type theory for invertibility in weak $ฯ‰$-categories

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ Homotopy Type Theory (HoTT) ์—์„œ๋Š” ๋ชจ๋“  ํƒ€์ž…์ด ์•ฝํ•œ ๊ณ ์ฐจ ๊ตฐ์ง‘(weak higher groupoid) ๊ตฌ์กฐ๋ฅผ ๊ฐ–๋Š”๋‹ค๋Š” ์‚ฌ์‹ค์ด ํ•ต์‹ฌ์ด๋‹ค. ์ด๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ Brunerie๋Š” ์•ฝํ•œ ฯ‰โ€‘๊ตฐ์ง‘์„ ํƒ€์ž… ์ด๋ก ์œผ๋กœ ๊ธฐ์ˆ ํ–ˆ๊ณ , Finsterโ€‘Mimram์€ ๋ฐฉํ–ฅ์„ฑ์„ ๋ถ€์—ฌํ•œ CaTT๋ฅผ ์ œ์‹œํ–ˆ๋‹ค. ์•ฝํ•œ ฯ‰โ€‘์นดํ…Œ๊ณ ๋ฆฌ์—์„œ ๊ฐ€์—ญ์„ฑ(invertibility) ์€ ๊ณ ์ฐจ ๋™ํ˜• ์‚ฌ์ƒ, ์•ฝํ•œ ๋™๋“ฑ์„ฑ, ๋ชจ๋ธ ๊ตฌ์กฐ ์ •์˜ ๋“ฑ์— ํ•„์ˆ˜์ ์ด์ง€๋งŒ, ์ฐจ์›์ด ๋ฌดํ•œํ•˜๊ธฐ์— ์ „ํ†ต์ ์ธ ๊ท€๋‚ฉ์  ์ •์˜๋Š” ๋ถˆ๊ฐ€๋Šฅํ•˜๊ณ  ๊ณต๋™๊ท€๋‚ฉ์  ์ ‘๊ทผ์ด ์š”๊ตฌ๋œ๋‹ค. 2.

Mathematics
No Image

A.E. Convergence vs Boundedness

1. ์—ฐ๊ตฌ ๋™๊ธฐ์™€ ๋ฐฐ๊ฒฝ ์„ ํ˜• ๊ฒฝ์šฐ : Stein(1970)์™€ Sawyer(1975)์˜ ๊ณ ์ „ ๊ฒฐ๊ณผ๋Š” โ€œ์„ ํ˜• ์—ฐ์‚ฐ์ž์˜ ์ ๋ณ„ ์ˆ˜๋ ด โ‡’ ์ตœ๋Œ€ํ•จ์ˆ˜์˜ ์•ฝํ˜• ์•ฝ์ •๋„โ€๋ฅผ ๋ณด์—ฌ์ค€๋‹ค. ์ด์ค‘(๋ฉ€ํ‹ฐ์„ ํ˜•) ๊ฒฝ์šฐ : ๊ธฐ์กด ๋ฌธํ—Œ์—์„œ๋Š” ๋ฐ˜๋Œ€๋กœ โ€œ์ตœ๋Œ€ํ•จ์ˆ˜์˜ ์•ฝํ˜• ์•ฝ์ •๋„ โ‡’ ์ ๋ณ„ ์ˆ˜๋ ดโ€์ด ์ฃผ๋กœ ๋‹ค๋ฃจ์–ด์กŒ์œผ๋ฉฐ, ์ˆ˜๋ ด์ด ์•ฝ์ •๋„๋ฅผ ๊ฐ•์ œํ•œ๋‹ค๋Š” ์—ญ๋ฐฉํ–ฅ์€ ๊ฑฐ์˜ ์•Œ๋ ค์ง€์ง€ ์•Š์•˜๋‹ค. ๋ณธ ๋…ผ๋ฌธ์˜ ๊ธฐ์—ฌ : ์ด ์—ญ๋ฐฉํ–ฅ์„ ์ด์ค‘ ์—ฐ์‚ฐ์ž์— ๋Œ€ํ•ด ์ตœ์ดˆ๋กœ ํ™•๋ฆฝํ•จ์œผ๋กœ์จ, ๋ฉ€ํ‹ฐ์„ ํ˜• ๋ถ„์„์—์„œ โ€œ์ˆ˜๋ ด โ‡’ ๊ฒฝ๊ณ„โ€๋ผ๋Š” ์ƒˆ๋กœ์šด ์—ฐ๊ฒฐ ๊ณ ๋ฆฌ๋ฅผ ์ œ๊ณตํ•œ๋‹ค. 2. ์ฃผ์š” ์ •๋ฆฌ์™€ ๊ทธ ์˜๋ฏธ | ์ •๋ฆฌ | ๋‚ด์šฉ | ์˜๋ฏธ |

Mathematics
No Image

Accelerated Discovery of Cryoprotectant Cocktails via Multi-Objective Bayesian Optimization

1. ์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋ฌธ์ œ ์ •์˜ ํฌ๋ผ์ด์˜ค๋ณด์กด์˜ ํ•ต์‹ฌ ๊ณผ์ œ : ๊ณ ๋†๋„ CPA๋Š” ๋น™๊ฒฐ์„ ์–ต์ œํ•ด ๋น„์ •์งˆ ๋™๊ฒฐ(vitrification)์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜์ง€๋งŒ, ๋†๋„๊ฐ€ ๋†’์„์ˆ˜๋ก ์„ธํฌ ๋…์„ฑ์ด ๊ธ‰์ฆํ•œ๋‹ค. ๋‹ค์ค‘๋ชฉํ‘œ ์„ค๊ณ„ ๊ณต๊ฐ„ : โ€œ๋†๋„(๋˜๋Š” vitrification ํšจ์œจ)โ€์™€ โ€œ์„ธํฌ ์ƒ์กด์œจโ€์ด๋ผ๋Š” ๋‘ ๋ชฉํ‘œ๊ฐ€ ์„œ๋กœ ์ƒ์ถฉํ•ด ์ „ํ†ต์ ์ธ ๋‹จ์ผ๋ชฉํ‘œ ์ตœ์ ํ™”๋‚˜ ๊ฒฝํ—˜์  ์ง๊ด€์— ์˜์กดํ•œ ํƒ์ƒ‰์€ ๋น„ํšจ์œจ์ ์ด๋‹ค. ์กฐํ•ฉ ํญ๋ฐœ : 7์ข… ์ด์ƒ์˜ CPA๋ฅผ 0.1 M ๋‹จ์œ„๋กœ ์กฐํ•ฉํ•˜๋ฉด ์ˆ˜์ฒœ~์ˆ˜๋งŒ ๊ฐœ์˜ ํ›„๋ณด๊ฐ€ ์ƒ์„ฑ๋ผ ์ „ํ†ต์ ์ธ ์ „์ˆ˜์กฐ์‚ฌ(exhaustive search)๋Š” ํ˜„์‹ค์ ์ด์ง€ ์•Š๋‹ค.

Computer Science Machine Learning

< Category Statistics (Total: 2829) >

Electrical Engineering and Systems Science
100
General
731
General Relativity
22
HEP-EX
17
HEP-LAT
3
HEP-PH
39
HEP-TH
19
MATH-PH
36
NUCL-EX
2
NUCL-TH
5
Quantum Physics
41

Start searching

Enter keywords to search articles

โ†‘โ†“
โ†ต
ESC
โŒ˜K Shortcut