ๆญฏ000000035964.PDF



Similar documents
ๆญฏ PDF

cat_data3.PDF

์ž๊ธฐ๊ตฌ์„ฑ์ง€๋„ ๊ธฐ๋ฐ˜ ๋ฐฉ๋ฒ•์„ ์ด์šฉํ•œ ์ด์ƒ ํƒ์ง€(Novelty Detection using SOM SOM-based Methods)

ๆญฏ522๋ฐ•๋ณ‘ํ˜ธ.PDF

10๊น€๋ฌ˜์„ 

R์„ ์ด์šฉํ•œ ํ…์ŠคํŠธ ๊ฐ์ •๋ถ„์„

<3131BFF92D3828C6D0B3CEBFACB1B82DC0CCBBF3C8A D38302E687770>

๋ฌด์„ ๋ฐ์ดํ„ฐ_์š”๊ธˆ์ œ์˜_๊ฐ€๊ฒฉ์ฐจ๋ณ„ํ™”์—_๊ด€ํ•œ_์—ฐ๊ตฌv4.hwp

DBPIA-NURIMEDIA

์ œ 1 ๋ถ€ ์—ฐ๊ตฌ ๊ฐœ์š”

<4D F736F F D20B1E2C8B9BDC3B8AEC1EE2DC0E5C7F5>

ยฑรจยผยบรƒยถ รƒรขยทร‚-1

ร‡รยถรณร€ร“ยปรงยบยธ7/8ยฟรน-ยพร•32

09๊ถŒ์˜ค์„ค_ok.hwp

์ž๊ธฐ๊ณต๋ช… ์ž„ํ”ผ๋˜์Šค ๋‹จ์ธต์ดฌ์˜ ๊ธฐ์ˆ  ์—ฐ๊ตฌ์„ผํ„ฐ \(MREIT Research Center\)

์‚ฌ์—…๋‹จ์†Œ์‹์ง€7ํ˜ธ

<35BFCFBCBA2E687770>

<C7D1B1B9B0E6C1A6BFACB1B8C7D0C8B828C0CCC1BEBFF85FC0CCBBF3B5B75FBDC5B1E2B9E9292E687770>

nonpara6.PDF

<31B1E8C1A4B7E6B9DAC1F6BCF6B1E8B9CCBCF72E687770>

์ •๋ณด๊ธฐ์ˆ ์‘์šฉํ•™ํšŒ ๋ฐœํ‘œ

High Resolution Disparity Map Generation Using TOF Depth Camera In this paper, we propose a high-resolution disparity map generation method using a lo

ๆญฏProduct1.PDF

untitled

<C5F0B0E82D313132C8A328C0DBBEF7BFEB292E687770>

DBPIA-NURIMEDIA

untitled

<32332D322D303120B9E6BFB5BCAE20C0CCB5BFC1D6312D32302E687770>

์ œ19๊ถŒ ์ œ3ํ˜ธ โ… . ๋ฌธ์ œ์ œ๊ธฐ ์˜จ๋ผ์ธ์„ ํ™œ์šฉํ•œ ๋‰ด์Šค ์„œ๋น„์Šค ์ด์šฉ์€ ์ด์ œ ๋” ์ด ์ƒ ์ƒˆ๋กœ์šด ์ผ์ด ์•„๋‹ˆ๋‹ค. ๋‰ด์Šค ์„œ๋น„์Šค๋Š” ์ด๋ฏธ ๊ธฐ์กด์˜ ์–ธ๋ก ์‚ฌ๋“ค์ด ๊ฐœ์„คํ•œ ์›น์‚ฌ์ดํŠธ๋ฅผ ํ†ตํ•ด ์ด๋ฃจ์–ด์ง€๊ณ  ์žˆ์œผ ๋ฉฐ ๊ธฐ์กด์˜ ์ข…์ด์‹ ๋ฌธ๊ณผ ๋ฐฉ์†ก์„ ์ œ์ž‘ํ•˜๋Š” ์–ธ๋ก ์‚ฌ๋“ค ์™ธ ์— ์˜จ๋ผ์ธ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•˜๋Š” ์‹ ์ƒ ์–ธ๋ก ์‚ฌ

E-IC-D-1-065(์ˆ˜์ •).hwp

< FC3BBBCD2B3E2C1A4C3A5C0C720C8AEC0E5B0FA20B9DFC0FC28C3D6C1BE31292E687770>

Ch 1 ๋จธ์‹ ๋Ÿฌ๋‹ ๊ฐœ์š”.pptx

DBPIA-NURIMEDIA

Software Requirrment Analysis๋ฅผ ์œ„ํ•œ ์ •๋ณด ๊ฒ€์ƒ‰ ๊ธฐ์ˆ ์˜ ์‘์šฉ

์œ„ํ•ด ์‚ฌ์šฉ๋œ ๊ธฐ๋ฒ•์— ๋Œ€ํ•ด ์†Œ๊ฐœํ•˜๊ณ ์ž ํ•œ๋‹ค. ์‹œ๊ฐํ™”์™€ ์ž๋ฃŒ๊ตฌ์กฐ๋ฅผ ๋™์‹œ์— ํ™œ์šฉํ•˜๋Š” ํ”„๋กœ๊ทธ๋žจ์ด ๊ฐ€์ง€๋Š” ํ•œ๊ณ„์™€ ์ด๋ฅผ ๊ทน๋ณตํ•˜๊ธฐ ์œ„ํ•œ ์‹œ๋„๋“ค์„ ์‚ดํŽด๋ด„์œผ๋กœ์„œ ์†Œ์…œ๋„คํŠธ์›Œํฌ์˜ ๋ถ„์„์„ ์œ„ํ•œ ์ ‘๊ทผ ๋ฐฉ์•ˆ์„ ๊ณ ์ฐฐํ•ด ๋ณด๊ณ ์ž ํ•œ๋‹ค. 2์žฅ์—์„œ๋Š” ์‹คํ—˜์— ์‚ฌ์šฉ๋œ ์ธํ„ฐ๋„ท ์ปค๋ฎค๋‹ˆํ‹ฐ์ธ MLBPark ๊ฒŒ์‹œํŒ

ร€รŒรร–รˆรฑ.hwp

<31362DB1E8C7FDBFF82DC0FABFB9BBEA20B5B6B8B3BFB5C8ADC0C720B1B8C0FC20B8B6C4C9C6C32E687770>

09์˜ค์ถฉ์›(613~623)

<BBE7B8B3B4EBC7D0B0A8BBE7B9E9BCAD28C1F8C2A5C3D6C1BE E687770>

DBPIA-NURIMEDIA

untitled

Microsoft Word - 04 _ __262 ๋ฐ•์žฅํ˜„

Microsoft Word - 15__ _ ์ž„๋•์›

ยผยบยฟรธรรธ รƒรขยทร‚-1

รยถยดรถรˆรฑ_0304_final.hwp

<313120C0AFC0FCC0DA5FBECBB0EDB8AEC1F2C0BB5FC0CCBFEBC7D15FB1E8C0BAC5C25FBCF6C1A42E687770>

Journal of Educational Innovation Research 2019, Vol. 29, No. 1, pp DOI: An Exploratory Stud

No Title

ๆญฏki ์กฐ์ค€๋ชจ.hwp

118 ๊น€์ •๋ฏผ ์†ก์‹ ์ฒ  ์‹ฌ๊ทœ์ฒ  ์„ ๋ฏธ์น˜๊ธฐ ๋•Œ๋ฌธ์ด๋‹ค(๊ฐ•์„์ง„ ๋“ฑ, 2000; ์‹ฌ๊ทœ์ฒ  ๋“ฑ, 2001; ์œค์น˜์› ๋“ฑ, 2005; ํ•˜ํƒœ๊ฒฝ ๋“ฑ, 2004; Schibeci, 1983). ๋ชจ๋‘  ๋‚ด์—์„œ ๊ตฌ์„ฑ์›๋“ค์ด ๊ณต๋™์œผ ๋กœ ์ถ”๊ตฌํ•˜๋Š” ํ•™์Šต ๋ชฉํ‘œ์˜ ๋‹ฌ์„ฑ์„ ์œ„ํ•˜์—ฌ ๊ฐ์ž ๋งก์€ ์—ญํ• ์— ๋”ฐ๋ผ ํ•จ๊ป˜

Microsoft Word - KSR2012A062.doc

:,,.,. 456, 253 ( 89, 164 ), 203 ( 44, 159 ). Cronbach ฮฑ= ,.,,..,,,.,. :,, ( )

DBPIA-NURIMEDIA

๋ฆฌ์ฝœ ๋ชจ๋“œ 1. ๋ฅผ ๋ˆŒ๋Ÿฌ ํฌ๋กœ๋…ธ๊ทธ๋ž˜ํ”„๊ฐ€ ์ค‘์ง€ํ–ˆ์„ ๋•Œ์˜ ๋ฉ”๋ชจ๋ฆฌ๋ฅผ ๋ฆฌ์ฝœํ•˜์‹ญ์‹œ์˜ค. 2. ๋ฅผ ๋ˆŒ๋Ÿฌ ๋‹ค์Œ ๋žฉ์„ ๋ณด์‹ญ์‹œ์˜ค. 3. ๋˜๋Š” ๋ฅผ ๋ˆŒ๋Ÿฌ ๋ฆฌ์ฝœ ๋ชจ๋“œ๋ฅผ ์ข…๋ฃŒํ•˜๊ณ  ํฌ๋กœ๋…ธ๊ทธ๋ž˜ํ”„ ๋ชจ๋“œ๋กœ ๋Œ์•„๊ฐ€์‹ญ์‹œ์˜ค. ํƒ€์ด๋จธ ๋ชจ๋“œ 1. ๋ฅผ ๋ˆŒ๋Ÿฌ ํƒ€์ด๋จธ ๋ชจ๋“œ๋กœ ์ „ํ™˜ํ•˜์‹ญ์‹œ์˜ค. 2. ๋ฅผ ๋ˆŒ๋Ÿฌ ํƒ€์ด๋จธ๋ฅผ ์‹œ

A Study on Forest Policy in Korea by Imperial Japan - With an Emphasis on the National Forest Policy

6.24-9๋…„ 6์›”

DBPIA-NURIMEDIA

<3130BAB9BDC428BCF6C1A4292E687770>

210_01

์•„ํƒœ์—ฐ๊ตฌ(์†ก์„์›) hwp

???? 1

ecorp-ํ”„๋กœ์ ํŠธ์ œ์•ˆ์„œ์ž‘์„ฑ์‹ค๋ฌด(์–‘์‹3)

ๆญฏ4์ฐจํ•™์ˆ ๋Œ€ํšŒ์›๊ณ (์žฅ์ง€์—ฐ).PDF

< C7CFB9DDB1E22028C6EDC1FD292E687770>

์ƒˆ๋งŒ๊ธˆ์„ธ๋ฏธ๋‚˜-1101-์ด์–‘์žฌ.hwp

untitled

๊ณตํ•™๋ฐ•์‚ฌํ•™์œ„ ๋…ผ๋ฌธ ์šด์˜ ์ค‘ ํ„ฐ๋„ํ™•๋Œ€ ๊ตด์ฐฉ์‹œ ์ง€๋ฐ˜๊ฑฐ๋™ ํŠน์„ฑ๋ถ„์„ ๋ฐ ํ”„๋กœํ…ํ„ฐ ์„ค๊ณ„ Ground Behavior Analysis and Protector Design during the Enlargement of a Tunnel in Operation 2011๋…„ 2์›” ์ธํ•˜๋Œ€

ร‡รยถรณร€ร“ยปรงยบยธ11/12ยฟรนรˆยฃ

๋Œ€๊ตฌ์ „์‹œ์ปจ๋ฒค์…˜์„ผํ„ฐ ์ „์‹œํ–‰์‚ฌ์˜ ์ง€์—ญ๊ฒฝ์ œ ํŒŒ๊ธ‰ํšจ๊ณผ ๋ถ„์„

<333820B1E8C8AFBFEB2D5A B8A620C0CCBFEBC7D120BDC7BFDC20C0A7C4A1C3DFC1A42E687770>

Journal of Educational Innovation Research 2019, Vol. 29, No. 2, pp DOI: 3 * Effects of 9th

VOL /2 Technical SmartPlant Materials - Document Management SmartPlant Materials์—์„œ ๊ธฐ๋ณธ์ ์ธ Document๋ฅผ ๊ด€๋ฆฌํ•˜๊ณ ์ž ํ•  ๋•Œ ํ•„์š”ํ•œ ์„ธํŒ…, ํŒŒ์ผ ์—…๋กœ๋“œ ๋ฐฉ๋ฒ• ๊ทธ๋ฆฌ๊ณ  Path Type์ธ Ph

<C0DBBEF7C1DF202D20C7D1B1B9BFA9BCBAC0CEB1C7C1F8C8EFBFF85FBFA9BCBAB0FA20C0CEB1C728C5EBB1C736C8A3292DB3BBC1F62E687770>

ร€ยฑยฝร‚ยฟรญ รƒรขยทร‚

<C7A5C1F620BEE7BDC4>

<91E6308FCD5F96DA8E9F2E706466>

ยฐรญยผยฎรร– รƒรขยทร‚

ํ˜„ ์•ˆ ๋ถ„ ์„ 2 Catsouphes & Smyer, 2006). ์šฐ๋ฆฌ๋‚˜๋ผ๋„ ์ˆ™๋ จ๋œ ์ธ ๋ ฅ๋ถ€์กฑ์— ๋Œ€ํ•œ ์šฐ๋ ค๊ฐ€ ์‹ฌํ™”๋˜๊ณ  ์žˆ์œผ๋ฉฐ, ์ผ์ž๋ฆฌ์˜ ๋ฏธ ์Šค๋งค์น˜ ์ˆ˜์ค€์ด ํ•ด์™ธ ์ฃผ์š”๊ตญ๋ณด๋‹ค ์‹ฌ๊ฐํ•˜๋‹ค๋Š” ์ ๋„ ์ง€ ์ง€๋ถ€์ง„ํ•œ ์œ ์—ฐ๊ทผ๋ฌด์ œ์˜ ํ™•์‚ฐ์„ ์œ„ํ•œ ์ง„์ •์„ฑ ์žˆ๋Š” ๋…ธ ๋ ฅ์ด ํ•„์š”ํ•˜๋‹ค๋Š” ์ ์„ ๋ณด์—ฌ์ค€๋‹ค

15 ํ™๋ณด๋‹ด๋‹น๊ด€ (์–ธ๋ก ํ™๋ณด๋‹ด๋‹น) ๊น€๋ณ‘ํ˜ธ ( ้‡‘ ็ง‰ ้Žฌ ) 16 (ํ–‰์ •๋‹ด๋‹น) ๋ฐ•์ฐฌํ•ด ( ๆœด ้‘ฝ ๆตท ) ์˜ˆ์‚ฐ๋‹ด๋‹น๊ด€ 17 (๋ณต์ง€ํ–‰์ •๋‹ด๋‹น) ์ดํ˜์žฌ ( ๏งก ่ตซ ๅœจ ) 18 (๋ณด์œก๋‹ด๋‹น) ์ฃผ์‚ฌ ์ด์˜์ž„ ( ๏งก ๆณณ ไปป ) ๊ธฐ๋™๊ทผ๋ฌดํ•ด์ œ. 19 (์žฅ์• ์ธ๋‹ด๋‹น) ๋ฐ•๋…ธํ˜ ( ๆœด ๏คน ็ˆ€ ) ๊ธฐ๋™

<4D F736F F D20B1E2C8B9BDC3B8AEC1EE2DC0E5BFEBC1D8>

์‚ผ๊ต-1-4.hwp

ํŠน์ง‘-5


์ •์ง„๋ช… ๋‚จ์žฌ์› ๋– ์˜ค๋ฅด๊ณ  ์žˆ๋‹ค. ๋ฐฐ๋‹ฌ์•ฑ์„œ๋น„์Šค๋Š” ์†Œ๋น„์ž๊ฐ€ ๋ฐฐ๋‹ฌ ์•ฑ์„œ๋น„์Šค๋ฅผ ์ด์šฉํ•˜์—ฌ ๋ฐฐ๋‹ฌ์Œ์‹์ ์„ ์ฐพ๊ณ  ์Œ์‹ ์„ ์ฃผ๋ฌธํ•˜๋ฉฐ, ๋Œ€๊ธˆ์„ ๊ฒฐ์ œ๊นŒ์ง€ ํ•  ์ˆ˜ ์žˆ๋Š” ์„œ๋น„ ์Šค๋ฅผ ๋งํ•œ๋‹ค. ๋ฐฐ๋‹ฌ์•ฑ์„œ๋น„์Šค๋Š” ๊ฐ„ํŽธํ•œ ์Œ์‹ ์ฃผ๋ฌธ ๊ณผ ๋ฐ”๋กœ๊ฒฐ์ œ ์„œ๋น„์Šค๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ ์ „ ์—ฐ๋ น์ธต์—์„œ ๋น ๋ฅด๊ฒŒ ๋ณด๊ธ‰๋˜๊ณ  ์žˆ๋Š” ๋ฐ˜๋ฉด,

WHO ์˜์ƒˆ๋กœ์šด๊ตญ์ œ์žฅ์• ๋ถ„๋ฅ˜ (ICF) ์—๋Œ€ํ•œ์ดํ•ด์™€๊ธฐ๋Šฅ์ ์žฅ์• ๊ฐœ๋…์˜ํ•„์š”์„ฑ ( ํ™ฉ์ˆ˜๊ฒฝ ) ๊Œ™ 127 ๋…ธ๋™์ •์ฑ…์—ฐ๊ตฌ ์ œ 4 ๊ถŒ์ œ 2 ํ˜ธ pp.127~148 c ํ•œ๊ตญ๋…ธ๋™์—ฐ๊ตฌ์› WHO ์˜์ƒˆ๋กœ์šด๊ตญ์ œ์žฅ์• ๋ถ„๋ฅ˜ (ICF) ์—๋Œ€ํ•œ์ดํ•ด์™€๊ธฐ๋Šฅ์ ์žฅ์• ๊ฐœ๋…์˜ํ•„์š”์„ฑํ™ฉ์ˆ˜๊ฒฝ *, (disabi

DBPIA-NURIMEDIA

WTVLMIYIJVCY.hwp

๋„๋น„๋ผ

<5BB9E8C0E7B4EBC7D0B1B35DBFACB1B8BAB8B0EDBCAD2DC3D6C1BEC3E2B7C22E687770>


์Šฌ๋ผ์ด๋“œ 1

ยตรฐร‡รƒ24-ร‡ยฅรรถยดรœยธรฉ

<31342DC0CCBFEBBDC42E687770>

:28 PM ํŽ˜์ด์ง€2 History of KOPEC Contents J a n u a r y KOPEC FAMILY 2005๋…„ 1์›”ํ˜ธ(ํ†ต๊ถŒ 276ํ˜ธ) ๋ฐœํ–‰์ผ 2005๋…„ 1์›” 15์ผ ๋“ฑ๋ก์ผ 1983๋…„ 7์›” 20์ผ ๋ฐœํ–‰์ธ

04ยฑรจยบรŽยผยบ

๊ฒฝํฌ 209_

44-4๋Œ€์ง€.07์ด์˜ํฌ532~

A sudy on realizaion of speech and speaker recogniion sysem based on feedback of recogniion value


45-51 ยนรšยผรธยธยธ

Transcription:

SVM

SVM 200112

200112

1 1 2 3 2.1 3 2.2 Labeled Unlabeled 8 2.2.1 8 2.2.2 9 2.2.3 Unlabeled 10 2.2.4 11 3 Unlabeled 12 3.1 Support Vector Machne 12 3.1.1 SVM 12 3.1.2 15 3.1.3 SVM 17 3.2 18 3.2.1 Unlabeled 19 3.2.2 Unlabeled 21 4 23 4.1 23 4.2 24 4.3 25

5 31 33 39

1. 4 2. Unlabeled 10 3. 2 SVM 12 4. Unlabeled 16 5. Unlabeled SVM 19 21 22 25 26 26 27 27 28 28 29 29 30 32

1. 20 2. 31

.,. 1960. 1980,.. k-,, Support Vector Machne,, Nave Bayes...,, unlabeled. unlabeled, unlabeled. labeled unlabeled. Support Vector Machne,, unlabeled,,

Reuter. Labeled, unlabeled., SVM, SVM. :, Support Vector Machne,,

ï

j w = j j j df N tf w log j tf j df d K w w w,...,, 2 1

2 χ

R D CSV f c : j d c j f c d CSV j f c d CSV τ j f c d CSV τ j d

ï

w ρ x ρ b = 0 x ρ w ρ b,, ξ ρ V w ρ b N ρ ρ 1 ρ ρ V w, b, ξ w w + C ξ 2 ξ ρ ρ ρ N 1 [ w x + b] ξ = : y 1 = 1 N = 1 : ξ > 0 W a ρ N N N ρ 1 ρ ρ W α = α + y yαα j x x j 2 = 1 = 1 j= 1 α ρ α α ρ N y α 0 1 : 0 α C = =

χ m c 1 } { = t log log k k k k k k t p c P c t P c t P t P c P c t P c t P t IG + =

log log, B A C A N A c P t P c t P c t MI + + = = c χ χ 2 2 D C B A D B C A CB AD N c t + + + + = χ χ χ χ χ χ d c t P d c t P c t RS k k k + + = log t k c P

1 } 1 k k k k k c t P c t P c t P c t P c t OR = 2 2 Y X XY Joachms [Hers94]

θ 1 var θ I T θ 2 ; ln = θ θ θ θ x f E I

w ρ x ρ b = 0 x ρ x ρ w ρ x ρ b = 0 w ρ x ρ b = 0 dst x ρ x ρ dst x ρ

Gven labeled example set L={x 1, y 1..., x l,y l } and unlabeled example set U={x 1,..., x u } and test example set T={s 1, y 1,..., s s,y s }; Intalze SVM classfer f 0 by tranng wth L 0 =L, U 0 =U Do 1. Set margn t as margn of f t 2. Set = margn t * 2 3. Set dstx = dstance between 4. Set U add = {x 1,y 1,...,x t,y t xu t, y=f t x, dstx } 5. Set L +1 = 6. Set U t+1 =U t -U add 7. Tran f t+1 wth L t+1 8.Classfy test set T wth f t+1 9.Set PR t as classfcaton precson of T 10. Set t = t+1 Whle U add >0 & PR t >PR t-1 Store the fnal SVM classfer

c C A A + B A A + Re Pr Re Pr 1 2 2 + + = β β β F β β β F β β β

χ

χ

[ 93],,, 1993. [ 00], FAQ,, 2000. [Angl87] D. Anglun, Learnng regular sets from queres and counterexamples, Informaton and Computaton, 75, pp. 87-106, 1993. [Angl93] D. Anglun, L. Hellersten and M. Karpnsk, Learnng read-once formulas wth queres, Journal of the ACM, 40, pp. 185-210, 1993. [Balu99] S. Baluja, Probablstc modelng for face orentaton dscrmnaton: Learnng from labeled and unlabeled examples, Advances n Neural Informaton Processng Systems 11, pp. 854-860, 1999. [Blum97] A. L. Blum and P. Langley, Selecton of Relevant Features and Examples n Machne Learnng, Artfcal Intellgence, Vol. 97, pp. 245-271, 1997. [Cast96] V. Castell and T. Cover, The relatve value of labeled and unlabeled samples n pattern recognton wth an unknown mxng parameter, IEEE Transactons on Informaton Theory, 426, pp. 2101-2117, 1996. [Cohn94] D. Cohn, L. Atlas and R. Landner, Improvng generalzaton wth actve learnng, Machne Learnng, 152, pp. 201-221, 1994. [Cohn96] D. Cohn, Z. Ghahraman and M. Jordan, Actve learnng wth statstcal models, Journal of Artfcal Intellgence Research, 4, pp. 29-145, 1996. [Crav98] M. Craven, D. Fretag, A. McCallum, T. Mtchell, K. Ngam and C. Quek, Learnng to extract symbolc knowledge from the World Wde Web, Techncal Report, School of Computer Scence, CMU, 1998. [Crav00] M. Craven, D. DPasquo, D. Fretag, A. McCallum, T. Mtchell, K. Ngam, and S. Slattery, Learnng to construct knowledge bases from the World Wde Web, Artfcal Intellgence, 1181-2, pp. 69-113, 2000.

[Day69] N. Day, Estmatng the components of a mxture of normal dstrbutons, Bometrka, 563, pp. 463-474, 1969. [Demp77] A. Dempster, N. Lard and D. Rubn, Maxmum lkelhood from ncomplete data va the EM algorthm, Journal of the Royal Statstcal Socety, Seres B, 391, pp. 1-38, 1977 [Dyer89] M. Dyer, A. Freze and R. Kannan, A random polynomal tme algorthm for approxmatng the volume of convex bodes, Proceedngs of the Annual ACM Symposum on the Theory of Computng, pp. 375-381, 1989. [Ghah94] Z. Ghahraman and M. Jordan, Supervsed learnng from ncomplete data va an EM approach. Advances n Neural Informaton Processng Systems 6, pp.120-127, 1994. [Gane89] S. Ganesalngam, Classfcaton and mxture approaches to clusterng va maxmum lkelhood, Appled Statstcs, 383, pp. 455-466, 1989. [Gane78] S. Ganesalngam and G. McLachlan, "The effcency of a lnear dscrmnant functon based on unclassfed ntal samples," Bometrka, 65, pp. 658-662, 1978 [Gros91] K. P. Gross, Concept acquston through attrbute evoluton and experment selecton, Doctoral dssertaton, School of Computer Scence, Carnege Mellon Unversty, Pttsburgh, PA., 1991. [Gud97] V. N. Gudvada, et.al, "Informaton Retreval on the World Wde Web," IEEE Internet Computng, Vol. 1, no. 5, September/October, 1997. [Hart98] H. Hardley and J. Rao, Classfcaton and estmaton n analyss of varance problems, Revew of Internatonal Statstcal Insttute, 36, pp. 141-147, 1968. [Hers94] W. R. Hersh, C. Buckley, T. J. Leone and D. H. Hckam, OHSUMED: An nteractve retreval evaluaton and new large test collecton for research, In Proceedngs of the 17th Annual ACM SIGIR Conference, pp. 192-201, 1994 [Jaak00] T. Jaakkola, M. Mela and T. Jebara, Maxmum entropy dscrmnaton, Advances n Neural Informaton Processng Systems 12, pp. 470-476, 2000.

[Joac97] T. Joachms, Text categorzaton wth support vector machnes: Learnng wth many relevant features, Techncal Report 23, Unverstät Dortmund, LS VIII, 1997. [Joac98] T.Joachms, Text Categorzaton wth Support Vector Machnes: Learnng wth Many Relevant Features, In Machne Learnng: ECML-98, Tenth European Conference on Machne Learnng, pp. 137-142, 1998 [Joac00] T. Joachms, Estmatng the Generalzaton Performance of an SVM Effcently, Proceedngs of the 17th Internatonal Conference on Machne Learnng ICML 2000, pp. 431-438, 2000. [Knob77] B. Knobe and K. Knobe, A method for nferrng context-free grammars, Informaton and Control, 31, pp.129-146, 1977. [Kwok98] J. T.-Y. Kwok, Automated text categorzaton usng support vector machne, In Proceedngs of the Internatonal Conference on Neural Informaton Processng, Ktakyushu, Japan, Oct, pp. 347-351, 1998. [Lang95] K. Lang, Newsweeder: Learnng to flter netnews, In Internatonal Conference on Machne Learnng ICML, pp. 331-339, 1995. [Lew94] D. Lews and Gale, A sequental algorthm for tranng text classfers, Proceedngs of ACM SIGIR Conference, 1994. [Lew97] D. Lews and K. Knowles, Threadng electronc mal: A prelmnary study, Informaton Processng and Management 332, pp. 209-207, 1997. [Ltt77] R. Lttle, Dscusson on the paper by Professor Dempster, Professor Lard and Dr. Rubn, Journal of the Royal Statstcal Socety, Seres B, 391, pp. 25, 1977 [Lova92] L. Lovasz and M. Smonovts, On the randomzed complexty of volume and dameter, Proceedngs of the IEEE Symposum on Foundatons of Computer Scence, IEEE, pp. 482-492, 1992. [Mcla75] G. McLachlan, Iteratve reclassfcaton procedure for constructng an asymptotcally optmal rule of allocaton n dscrmnant analyss, Jounal of the

Amercal Statstcal Assocaton, 79350, pp. 365-369, 1975. [Mcla82] G. McLachlan, Updatng a dscrmnant functon on the bass of unclassfed data, Communcatons n Statstcs: Smulaton and Computaton, 116, pp. 753-767, 1982. [Mere00] D. Meretaks, D. Fragouds, H. Lu and S. Lkothanasss, Scalable Assocatonbased Text Classfcaton, Proceedngs of CIKM-00, 9th ACM Internatonal Conference on Informaton and Knowledge Management,pp. 5-11, Washngton, US,2000. [Mll96] D. Mller and H. Uyar, A generalzed Gaussan mxture classfer wth learnng based on both labelled and unlabelled data, Proceedngs of the 1996 Conference on Informaton Scence and Systems. 1996. [Moon00] R. Mooney and L. Roy, Context-based book recommendng usng learnng for text categorzaton, In Proceedngs of the Ffth ACM Conference on Dgtal Lbrares, pp.195-204, 2000. [Murr78] G. Murray and D. Ttterngton, Estmaton problems wth data from a mxture, Appled Statstcs, 273, pp. 325-334, 1978. [Nga98] K. Ngam, A. McCallum, S. Thrun and T. Mtchell, Learnng to Classfy Text from Labeled and Unlabeled Documents, In Proceedngs of AAAI-98, 15th Conference of the Amercan Assocaton for Artfcal Intellgence, pp. 792-799, 1998. [One78] T. O'Nell, Normal dscrmnaton wth unclassfed observatons, Journal of the Amercan Statstcal Assocaton, 73364, pp. 821-826, 1978. [Pazz96] M. Pazzan and D. Bllsus, Syskll & Webert: Identfyng nterestng Web stes, In Proceedngs of the Thrteenth Natonal Conference on Artfcal Intellgence,pp. 54-61, Portland, 1996. [Rats95] J. Ratsaby and S. Venkatesh, Learnng from a mxture of labeled and unlabeled examples wth parametrc sde nformaton, Proceedngs of the Eghth Annual Conference on Computatonal Learnng Theory, pp. 412-417, 1995.

[Rve93] R. L. Rvest, and R. E. Schapre, Inference of fnte automata usng homng sequences, Informaton and Computaton, 103, pp. 299-347, 1993. [Samm86] C. Sammut and R.B. Banerj, Learnng concepts by askng questons, In R.S. Mchalsk, J.G. Carbonell, & T.M. Mtchell Eds., Machne learnng: An artfcal ntellgence approach Vol.2. San Francsco, CA: Morgan Kaufmann, 1986. [Seba99] F. Sebastan, "Machne Learnng n Automated Text Categorsaton," Techncal Report IEI-B4-31-1999, Isttuto d Elaborazone dell'informazone, Consglo Nazonale delle Rcerche, Psa, IT, 1999. [Seun92] H. Seung, M. Opper and H. Sompolnsky, Query by Commttee, Proceedngs of the Ffth Annual Workshop on Computatonal Learnng Theory, pp. 287-294, New York, ACM Press, 1992. [Shah94] B. Shahshahan and D. Landgrebe, The effect of unlabeled samples n reducng the small sze problem and mtgatng the Huges phenomenon, IEEE Transactons on Geoscence and Remote Sensng, 325, pp. 1087-1095, 1994. [Shav98] J. Shavlk and T. Elass-Rad, Intellgent agents for web-based tasks: An advce-takng approach, Learnng for Text Categorzaton: Papers from the AAAI Workshop, pp. 63-70. Tech. rep. WS-98-05, AAAI Press. 1998. [Snc89] A. Snclar and M. Jerrum, Approxmate countng, unform generaton and rapdly mxng Markov chans, Informaton and Computaton, 82, pp. 93-133, 1998. [Szum01] M. Szummer and T. Jaakkola, Kernel expansons wth unlabeled data, Advances n Neural Informaton Processng Systems 13, 2001. [Ttt76] D. Ttterngton, Updatng a dagnostc system usng unconfrmed cases, Appled Statstcs, 253, pp. 238-247, 1976. [Vapn95] V. Vanpk, The Nature of Statstcal Learnng Theory, Sprnger-Verlag, 1995 [Yang99a] Y. Yang, et al., "Learnng Approaches for Detectng and Trackng News Events," IEEE Intellgent System, pp. 32-43, July/August 1999.

[Yang99b] Y. Yang and X. Lu, "A Re-examnaton of Text Categorzaton Methods," Proceedngs of the 22h Annual Internatonal ACM SIGIR Conference on Research and Development n Informaton Retreval SIGIR 99, pp. 42-49, 1999. [Yang99c] Y. Yang, "An Evaluaton of Statstcal Approaches to Text Categorzaton," Journal of Informaton Retreval, vol 1, no. 1/2, pp. 67-88, 1999. [Zhan00] T. Zhang and F. Oles, A probablty analyss on the value of unlabeled data for clasfcaton problems, Proceedngs of the Seventeenth Internatonal Conference on Machne Learnng, pp. 1191-1198, 2000.

ABSTRACT Incremental Supervsed Learnng based on SVM wth Unlabeled Documents Soo-Young Km Dept. of Computer Scence The Graduate School Yonse Unversty

Keywords : Unlabeled data, Support Vector Machne, Incremental Supervsed Learnng, Text Categorzaton