๊ด€๋ฆฌ ๋ฉ”๋‰ด

Done is Better Than Perfect

01. introduction ๋ณธ๋ฌธ

๐Ÿค– AI/Machine Learning

01. introduction

jimingee 2022. 1. 20. 16:42

 

    1.  Machine Learning์˜ ์ •์˜

    • [Arthur Samuel์˜ ์ •์˜] ํ”„๋กœ๊ทธ๋ž˜๋ฐ ์—†์ด๋„ ์ปดํ“จํ„ฐ ์Šค์Šค๋กœ ํ•™์Šตํ•  ์ˆ˜ ์žˆ๊ฒŒ ๋งŒ๋“ค์–ด ์ฃผ๋Š” ์—ฐ๊ตฌ
    " Machine Learning is a field of study that gives computers the ability to learn without being explicitly programmed."

     

    • [Tom Mitchell์˜ ์ •์˜] ์ž‘์—… T๊ฐ€ ์žˆ๊ณ , P๋กœ ์„ฑ๋Šฅ์ด ์ธก์ •๋  ๋•Œ, ๊ทธ P๊ฐ€ ๊ฒฝํ—˜ E๋ฅผ ํ†ตํ•ด ํ•™์Šตํ•˜๋ฉฐ ํ–ฅ์ƒ๋˜๋Š” ํ”„๋กœ๊ทธ๋žจ
    Tom Mitchell provides a more modern definition: 
    "A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.

    Example: playing checkers.

    E = the experience of playing many games of checkers

    T = the task of playing checkers.

    P = the probability that the program will win the next game.

     

    ์‘์šฉ๋ถ„์•ผ

    • data mining
    • ์†์œผ๋กœ ํ”„๋กœ๊ทธ๋ž˜๋ฐํ•  ์ˆ˜ ์—†๋Š” ๋ถ„์•ผ (์ž์œจ์ฃผํ–‰ ์ž๋™์ฐจ, ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ, ์˜์ƒ์ฒ˜๋ฆฌ)
    • Self-customizing programs
    • understanding human learning(brain, real AI)

     

     

    2.  Machine Learning ์•Œ๊ณ ๋ฆฌ์ฆ˜

    • Supervised Learning(์ง€๋„ ํ•™์Šต) : ์ •๋‹ต์„ ์•Œ๊ณ  ์žˆ๋Š” ๋ฐ์ดํ„ฐ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์˜ˆ์ธก ๋ชจ๋ธ์„ ๋งŒ๋“œ๋Š” ML
    • Unsupervised Learning(๋น„์ง€๋„ ํ•™์Šต) :  ์ •๋‹ต์„ ๋ชจ๋ฅด๋Š” ๋ฐ์ดํ„ฐ์—์„œ ์œ ์šฉํ•œ ์ •๋ณด๋ฅผ ๋ถ„๋ฅ˜, ์ถ”์ถœํ•˜๋Š” ML
    • ์ถ”๊ฐ€์ ์œผ๋กœ ์‚ฌ์šฉ์ž๊ฐ€ ๊ด€์‹ฌ์„ ๊ฐ€์งˆ๋งŒํ•œ ์ •๋ณด๋ฅผ ์ถ”์ฒœํ•˜๋Š” Recommender System๋„ ML์˜ ํ•œ ๋ถ„์•ผ์ด๋‹ค. 

     

     

     

    3.  Supervised Learning  (์ง€๋„ ํ•™์Šต)

    ์ •๋‹ต(label)์ด ์žˆ๋Š” dataset์„ ์‚ฌ์šฉํ•˜์—ฌ ์ปดํ“จํ„ฐ๋ฅผ ํ•™์Šต์‹œํ‚ค๋Š” ๋ฐฉ๋ฒ•.  data set์˜ ํ˜•ํƒœ๋Š” [data(input) - label(output)]์œผ๋กœ ์ฃผ์–ด์ง„๋‹ค. 

    ๋”ฐ๋ผ์„œ label์ด ์ •ํ•ด์ง€์ง€ ์•Š์€ data๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ label์„ ์˜ˆ์ธกํ•œ๋‹ค. 

    Supervised Learning ์˜ ์„ธ๋ถ€ ๋ถ„๋ฅ˜๋กœ๋Š” 'regression'๊ณผ 'classification'์ด ์žˆ๋‹ค. 


    regression : ์—ฐ์†์ ์ธ(continuous) ๊ฒฐ๊ณผ๊ฐ’ ์˜ˆ์ธก

    ex) ๋ถ€๋™์‚ฐ ์ง‘ ๊ฐ’ ์˜ˆ์ธก

    ๊ฐ•์˜์˜ ์˜ˆ์ œ์—์„œ input์€ ์ง‘์˜ ๋„“์ด์ด๊ณ , output์€ ์ง‘์˜ ๊ฐ€๊ฒฉ์ด๋‹ค. 

    output์— ํ•ด๋‹นํ•˜๋Š” ์ง‘์˜ ๊ฐ€๊ฒฉ์€ ์—ฐ์†์ ์ธ ๊ฐ’์„ ๊ฐ€์ง€๋ฏ€๋กœ 

    ์ด๋Š” regression ๋ฌธ์ œ์ด๋‹ค. 

     

     

     


    classification : ์ด์‚ฐ์ ์ธ(descrete) ๊ฒฐ๊ณผ๊ฐ’ ์˜ˆ์ธก

    ex ) ์ข…์–‘ ํฌ๊ธฐ์— ๋”ฐ๋ฅธ ์–‘์„ฑ/์Œ์„ฑ ์˜ˆ์ธก

    ๊ทธ๋ฆผ์˜ ์˜ˆ์ œ์—์„œ 

    input์€ ์ข…์–‘์˜ ํฌ๊ธฐ์ด๊ณ , output์ธ ์ง„๋‹จ๊ฒฐ๊ณผ์ด๋‹ค.

    output์— ํ•ด๋‹นํ•˜๋Š” ์ง„๋‹จ๊ฒฐ๊ณผ๊ฐ€ ์–‘์„ฑ/์Œ์„ฑ์œผ๋กœ discrete category์ด๋ฏ€๋กœ 

    ์ด๋Š” classification ๋ฌธ์ œ์ด๋‹ค.

     

     

     

    ์•„๋ž˜ ๊ทธ๋ž˜ํ”„์ฒ˜๋Ÿผ ์ข…์–‘์˜ ํฌ๊ธฐ์— ๋‚˜์ด ์†์„ฑ์„ ๋”ํ•˜์—ฌ ์ง„๋‹จ๊ฒฐ๊ณผ๋ฅผ ๋‚ด๋ฆด ์ˆ˜ ์žˆ๋‹ค.

    ์ถ”๊ฐ€์ ์œผ๋กœ Clump ๋‘๊ป˜, ์ข…์–‘ ์„ธํฌํฌ๊ธฐ์˜ ๊ท ์ผํ•จ, ๋ชจ์–‘์˜ ๊ท ์ผํ•จ ๋“ฑ ๋‹ค์–‘ํ•œ ์†์„ฑ(feature)๋„ ํ•จ๊ป˜ ๊ณ ๋ คํ•˜์—ฌ ๋ถ„๋ฅ˜ํ•  ์ˆ˜ ์žˆ๋‹ค.

    ์ด๋ ‡๋“ฏ ์†์„ฑ์˜ ๊ฐœ์ˆ˜๊ฐ€ ์ฆ๊ฐ€ํ• ์ˆ˜๋ก, ๊ทธ๋ž˜ํ”„์˜ ์ถ•(์ฐจ์›)์˜ ๊ฐœ์ˆ˜ ๋˜ํ•œ ์ฆ๊ฐ€ํ•œ๋‹ค.  ์ฆ๊ฐ€ํ•œ ์ฐจ์›์—์„œ ์„ ํ˜•์˜ ์ง์„ ์œผ๋กœ ๊ตฌ๋ถ„์ง“๋Š” ํ•จ์ˆ˜๋ฅผ ์ฐพ๋Š” ๊ฒƒ์ด ์˜ˆ์ธก ๋ชจ๋ธ์„ ์ฐพ๋Š” ๊ฒƒ์ด๋ผ ํ•  ์ˆ˜ ์žˆ๋‹ค.

     

     

    4.  Unsupervised Learning (๋น„์ง€๋„ ํ•™์Šต)

    ์ •๋‹ต(label)์ด ์—†๋Š” dataset์„ ์‚ฌ์šฉํ•˜์—ฌ ์ปดํ“จํ„ฐ๋ฅผ ํ•™์Šต์‹œํ‚ค๋Š” ๋ฐฉ๋ฒ•.  data set์˜ ํ˜•ํƒœ๋Š” [data(input)]์œผ๋กœ ์ฃผ์–ด์ง„๋‹ค.

    ๋”ฐ๋ผ์„œ ๋ฐ์ดํ„ฐ์— ์ˆจ๊ฒจ์ง„ ๊ตฌ์กฐ๋‚˜ ํŠน์ง•์„ ๋ฐœ๊ฒฌํ•˜๋Š” ๋ฐ ์‚ฌ์šฉ๋œ๋‹ค.

     

    • clustering : ์ฃผ์–ด์ง„ ๋ฐ์ดํ„ฐ๋ฅผ n๊ฐœ์˜ ์นดํ…Œ๊ณ ๋ฆฌ๋กœ ๊ตฐ์ง‘ํ™”
    • non-clustering : cocktail party problem - ์นตํ…Œ์ผ ํŒŒํ‹ฐ์žฅ ๋…น์Œ๋ฐ์ดํ„ฐ์—์„œ ํŠน์ • ์Œ์„ฑ๋งŒ ๋ถ„๋ฆฌํ•ด๋ƒ„

     

     


    ์ฐธ๊ณ  ๋ฌธํ—Œ

    machine learning ๊ฐ•์˜ ๋…ธํŠธ

     

     

    '๐Ÿค– AI > Machine Learning' ์นดํ…Œ๊ณ ๋ฆฌ์˜ ๋‹ค๋ฅธ ๊ธ€

    04. Logistic Regression  (1) 2022.03.23
    03. Linear Regression with Multiple Variable  (1) 2022.02.10
    02. Linear regression with one variable  (1) 2022.02.04
    Machine Learning ์ •๋ฆฌ  (0) 2022.01.20
    ์ถ”์ฒœ ์‹œ์Šคํ…œ์ด๋ž€  (0) 2022.01.05
    Comments