[Foundation of Supervised Learning] part 1 - 1

2023. 1. 12. 19:45

๐ŸŽฏ Keyword ๐ŸŽฏ

- Supervised Learning
- Learning pipeline
- Hypothesis f
- Learning model

 

 

์—ฌ๋Ÿฌ ๊ณ ์–‘์ด ์‚ฌ์ง„

์ด ๋™๋ฌผ์˜ ์ด๋ฆ„์€ ๋ฌด์—‡์ธ๊ฐ€์š”?

 

์—ฌ๋Ÿฌ๋ถ„์€ ์œ„ ์งˆ๋ฌธ์„ ๋ณด๊ณ  ๋ฐ”๋กœ ๊ท€์—ฌ์šด ๊ณ ์–‘์ด๋‹ค ! ๋ผ๊ณ  ๋งํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

 

์œ„ ๊ท€์—ฌ์šด ๊ณ ์–‘์ด๋ฅผ ๋ณด๊ณ  ์—ฌ๋Ÿฌ๋ถ„์€ ๋™๋ฌผํ•™์ ์ธ ์ •์˜๋กœ ๊ณ ์–‘์ด์ž…๋‹ˆ๋‹ค๋ผ๊ณ  ํ•˜๊ธฐ ๋ณด๋‹ค๋Š” ๊ฒฝํ—˜์ ์œผ๋กœ ์•Œ๊ฒŒ ๋˜์—ˆ๋‹ค๊ณ  ํ•  ๊ฒƒ์ž…๋‹ˆ๋‹ค.

 

Machine Learning๋„ ์ด์™€ ๊ฐ™์ด Data๋กœ๋ถ€ํ„ฐ ๋‚ด์žฌ๋œ ํŒจํ„ด์„ ํ•™์Šตํ•˜๋Š” ๊ณผ์ •์ž…๋‹ˆ๋‹ค.

 

image recognition problem, ์ง‘๊ฐ’์˜ ์ถ”์ด ๋ถ„์„, ํ†ต๊ณ„ ๋ถ„์„ ์€ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ๋ถ„๋ฅ˜ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

 

Machine Learning Problem

 

  • Binary Classification
  • Multiclass Classification
  • Regression

 

Machine Learning์€ ๋‹ค์‹œ

 

Supervised Learning๊ณผ Unsupervised Learning์œผ๋กœ ๋‚˜๋ˆŒ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

 

๋‘˜์˜ ํฐ ์ฐจ์ด๋Š” Supervised Learning์€ label์ด ์žˆ๋Š” ๊ฒƒ์ด๊ณ , Unsupervised Learning์€ laebl์ด ์—†๋Š” data์ธ ๊ฒƒ์ž…๋‹ˆ๋‹ค.

 

์š”์•ฝํ•˜๋ฉด ์•„๋ž˜์™€ ๊ฐ™์Šต๋‹ˆ๋‹ค.

 


                                                                                                    Regression     => Continuous output

                                            Supervised Learning ------->     Classification =>  Discrete output

                                            (Labeled Data)

Machine Learning - - - > 

                                            Unsupervised Learning ------->  Clustering

                                            (Unlabeled Data)                             Dimensional reduction


 

Supervised Learning์˜ Data Set์€ ์ž…๋ ฅ x์™€ ์ถœ๋ ฅ y์˜ ์Œ์œผ๋กœ ๊ตฌ์„ฑ์ด ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.

 

Supervised Learning์˜ ๋ชฉ์ 
"x -> y๊ฐ€๋Š” ํ•จ์ˆ˜ h๋ฅผ ํ•™์Šตํ•˜๋Š” ๊ฒƒ"

 

 

 

Learning pipeline์€ ๋‹ค์Œ๊ณผ ๊ฐ™์ด ์ด๋ฃจ์–ด์ง‘๋‹ˆ๋‹ค.

 

  • ์ดˆ๊ธฐ์—๋Š” Machine Learning model ์ด ์ž˜ ํ•™์Šตํ•˜์ง€ ๋ชปํ•จ.
  • ๊ทธ๋Ÿฌ๋‚˜ ์šฐ๋ฆฌ๊ฐ€ ๊ฐ€์ง€๊ณ  ์žˆ๋Š” Data set label๋กœ๋ถ€ํ„ฐ , ์šฐ๋ฆฌ๊ฐ€ ๊ฐ€์ง€๊ณ  ์žˆ๋Š” ์ •๋‹ต์œผ๋กœ๋ถ€ํ„ฐ ์ด ์ถœ๋ ฅ์„ ํ•™์Šตํ•จ.
  • model์˜ parameter ๊ฐ’์„ ๋ณ€๊ฒฝํ•ด ๋‚˜๊ฐ. 
    • Supervised Learning์—์„œ๋Š” model output๊ณผ ์ •๋‹ต๊ณผ์˜ ์ฐจ์ด์ธ Error๋ฅผ ํ†ตํ•ด์„œ error๋ฅผ ์ค„์—ฌ๊ฐ€๋ฉฐ ํ•™์Šตํ•จ.
    • ํ•™์Šต Sample์ด ์ฃผ์–ด์ง€๋ฏ€๋กœ ๊ฐ€๋Šฅํ•œ ๊ฒƒ.

 

 

Test ๋‹จ๊ณ„ (๋ชจ๋ธ์ด ํ™˜๊ฒฝ์— ์ ์‘ํ•˜๋Š”์ง€ ํ™•์ธ)

 

- > training๊ณผ ์ž…๋ ฅ Data ๋‹ค๋ฆ„.

 

 

 

Input Feature

-> ํ•ด๋‹น ๋ถ„์•ผ์˜ ์ „๋ฌธ ์ง€์‹์ด ์žˆ์–ด์•ผ classification์— ํšจ๊ณผ์ ์ธ feature๋ฅผ ๋„์ž…ํ•˜๊ธฐ ์–ด๋ ค์šธ ๊ฒƒ์ž…๋‹ˆ๋‹ค.

-> Input Feature์˜ Design์€ ๊ด€๋ จ ๋ถ„์•ผ์˜ ์ „๋ฌธ์„ฑ์„ ๊ฐ€์ง„ ์ „๋ฌธ๊ฐ€์™€ ํ•จ๊ป˜ ํ•ด๊ฒฐํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

 

   => ๊ทธ๋Ÿฌ๋‚˜ ์ตœ๊ทผ ๋”ฅ๋Ÿฌ๋‹ ๋ชจ๋ธ์€ ์ด๋Ÿฌํ•œ feature ์—ญ์‹œ ์Šค์Šค๋กœ ํ•™์Šตํ•˜๊ธฐ ๋•Œ๋ฌธ์— Domain knowledge๋กœ๋ถ€ํ„ฐ ๋น„๊ต์  ์ž์œ ๋กœ์šด ํŽธ์ž…๋‹ˆ๋‹ค.

 

 

 

Binary Classification

=> Output -> Yes or No

 

 

Target Function

-> ๋‹ฌ์„ฑํ•˜๊ธฐ ์–ด๋ ค์šด ์ด์ƒ์ ์ธ ํ•จ์ˆ˜, ๋ชจ๋ธ

=> ์ˆ˜๋งŽ์€ data๋ฅผ ํ†ตํ•ด target function์— ๊ทผ์ ‘ํ•œ ๋ชจ๋ธ์„ ๋งŒ๋“œ๋Š” ๊ฒƒ์ด ์šฐ๋ฆฌ๊ฐ€ ํ•  ์ˆ˜ ์žˆ๋Š” ์ผ์ž…๋‹ˆ๋‹ค.

=> ์ด Target Function์„ hypothesis H ๋ผ๊ณ  ์ •์˜ํ•˜๊ณ  ์šฐ๋ฆฌ๊ฐ€ ๊ฐ€์ง€๊ณ  ์žˆ๋Š” ๋ชจ๋ธ์ด ๋ฐ”๋กœ "Machine Learning"์ž…๋‹ˆ๋‹ค.

 

 

Target Function F ๋ฅผ ์œ„ํ•ด์„œ ๋‹ค์Œ๊ณผ ๊ฐ™์€ ๋‹จ๊ณ„๋ฅผ ๊ฑฐ์นฉ๋‹ˆ๋‹ค.

  • fearture selection
  • model selection
  • optimization

 

Model Selection

-> ์šฐ๋ฆฌ๊ฐ€ ํ’€๊ณ ์ž ํ•˜๋Š” ๋ฌธ์ œ์— ๊ฐ€์žฅ ์ ํ•ฉํ•œ ๋ชจ๋ธ์„ ์„ ์ •ํ•˜๋Š” ๊ฒƒ.

     -> ๊ทธ ์ค‘์—๋Š” ์„ ํ˜• ๋ชจ๋ธ๋„ ์กด์žฌํ•˜๊ณ  Neural Network์™€ ๊ฐ™์€ ๋น„์„ ํ˜• ๋ชจ๋ธ๋„ ์กด์žฌํ•ฉ๋‹ˆ๋‹ค.

 

Optimization

-> ๋ชจ๋ธ Parameter๋ฅผ ๋ฐ”๊พธ์–ด model์ด ๊ฐ€์žฅ ์šฐ์ˆ˜ํ•œ ์„ฑ๋Šฅ์„ ์ œ๊ณตํ•˜๋„๋ก ํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค.

'Artificial Intelligence' ์นดํ…Œ๊ณ ๋ฆฌ์˜ ๋‹ค๋ฅธ ๊ธ€

[Gradient Descent] part 2 - 3  (0) 2023.01.14
[Linear Regression] part 2 - 2  (0) 2023.01.14
[Linear Regression] part 2 - 1  (0) 2023.01.14
[Foundation of Supervised Learning] part 1 - 2  (0) 2023.01.12
Deep Learning ๊ธฐ๋ฐ˜ ์ดํ•ด  (0) 2022.02.21

BELATED ARTICLES

more