Upgrade to Pro — share decks privately, control downloads, hide ads and more …

머신러닝을 위한 기초 수학 살펴보기

mingrammer
August 11, 2017

머신러닝을 위한 기초 수학 살펴보기

머신러닝을 위한 매우 기초적인 수학과 이를 응용한 선형 회귀 학습 예제

mingrammer

August 11, 2017
Tweet

More Decks by mingrammer

Other Decks in Technology

Transcript

  1. ݠन۞׬ਸਤೠӝୡࣻ೟࢓ಝࠁӝ (FUUJOHTUBSUFEUPMFBSOUIFMJOFBSBMHFCSBXJUIQZUIPO ӂ޹੤

  2. ݠन۞׬ਸ ਤೠ ӝୡ ࣻ೟ ࢓ಝࠁӝ MinJae Kwon (@mingrammer) 2017.08.12 (Getting

    started to learn the linear algebra with python)
  3. Name ӂ޹੤ (MinJae Kwon) Nickname @mingrammer Email [email protected] Who Game

    Server Engineer @ SundayToz Blog https://mingrammer.com Facebook https://facebook.com/mingrammer Github https://github.com/mingrammer Eng Blog https://medium.com/@mingrammer
  4. 2. ࢶഋ؀ࣻ೟੉ۆ? 4. ୶о ӝୡ ࣻ೟ ࢓ಝࠁӝ Contents 5. ࢶഋ

    ഥӈ ҳഅ೧ࠁӝ 1. ࣻ೟੄ ೙ਃࢿ 3. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ 6. Next (more LA and Mathematics)
  5. ࣻ೟੄ ೙ਃࢿ ੉ߣࣁ࣌਷઱۽ӝୡ ࢶഋ؀ ࣻ೟ਸ׮ܖ޲۽ ׮਺ࣻधٜ੉੊ࣼೞ૑ঋ਷ٜ࠙ਸ؀࢚ਵ۽೤פ׮ W ←W −α ∂L

    ∂W E = − 1 N t nk log(y nk ) k ∑ n ∑ d(u ! u ! T ) = 2u !
  6. ࣻ೟੄ ೙ਃࢿ ਢ ѐߊ জ ѐߊ API ѐߊ ࣻ೟ T(x)

    ∂ ∂θ f (x,θ)dx ∫ −log(t)y(t) ∑ x∇f (x) 1 σ 2π e −(x−µ)2 2σ 2 −∞ ∞ ∫
  7. ࣻ೟੄ ೙ਃࢿ ਢ ѐߊ জ ѐߊ API ѐߊ ࣻ೟ *UEFQFOETPOjCVU

    T(x) ∂ ∂θ f (x,θ)dx ∫ −log(t)y(t) ∑ x∇f (x) 1 σ 2π e −(x−µ)2 2σ 2 −∞ ∞ ∫
  8. ࣻ೟੄ ೙ਃࢿ ߑޙ੗ࣻ҅ஏ ୶ୌঌҊ્ܻ ঐഐࢸ҅ ঑୷ঌҊ્ܻ %ݽ؛݂ ѱ੐ূ૓ ୭ࣗ࠺ਊঌҊ્ܻ ೐۽ࣁझझாે݂

    Ӓې೗୊ܻ ഛܫݽ؛ ҊബਯҊࢿמ҅࢑ࢸ҅ ݠन۞׬ ؘ੉ఠ߬੉झ नഐ୊ܻ
  9. ࣻ೟੄ ೙ਃࢿ ݠन۞׬ ߑޙ੗ࣻ҅ஏ ୶ୌঌҊ્ܻ ঐഐࢸ҅ ঑୷ঌҊ્ܻ %ݽ؛݂ ѱ੐ূ૓ ୭ࣗ࠺ਊঌҊ્ܻ

    ೐۽ࣁझझாે݂ Ӓې೗୊ܻ ഛܫݽ؛ ҊബਯҊࢿמ҅࢑ࢸ҅ ؘ੉ఠ߬੉झ नഐ୊ܻ
  10. ࣻ೟੄ ೙ਃࢿ .BDIJOF-FBSOJOH "MHPSJUINT 5SBJO%BUB .PEFMT 7BMJEBUJPO%BUB 0VUQVUT /FX%BUB

  11. ࣻ೟੄ ೙ਃࢿ 6 7 8 ... 5 7 5 6

    ... 3 8 4 1 ... 4 ... ... ... ... 2 0 1 5 4 3 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ W = W −α ∂L ∂W 0 0 1 ... 0 1 0 0 ... 0 0 1 0 ... 0 ... ... ... ... 0 0 0 0 0 1 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 0 1 0 ... 0 1 0 0 ... 0 0 1 0 ... 0 ... ... ... ... 0 0 0 1 0 0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 4 1 5 ... 8 ⎡ ⎣ ⎤ ⎦ 1 2 (y k − t k )2 ∑ y k = eak eai ∑ 0.5 0.4 ... ... 0.8 0.1 −0.3 ... ... 0.2 ... ... ... ... 0.1 ... ... ... ... 0.43 0.03 0.23 0.1 0.3 0.13 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 0 1 0 ... 0 ⎡ ⎣ ⎤ ⎦ Y = X ⋅W + B
  12. ࣻ೟੄ ೙ਃࢿ 6 7 8 ... 5 7 5 6

    ... 3 8 4 1 ... 4 ... ... ... ... 2 0 1 5 4 3 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ W = W −α ∂L ∂W 0 0 1 ... 0 1 0 0 ... 0 0 1 0 ... 0 ... ... ... ... 0 0 0 0 0 1 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 0 1 0 ... 0 1 0 0 ... 0 0 1 0 ... 0 ... ... ... ... 0 0 0 1 0 0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 4 1 5 ... 8 ⎡ ⎣ ⎤ ⎦ 1 2 (y k − t k )2 ∑ y k = eak eai ∑ 0.5 0.4 ... ... 0.8 0.1 −0.3 ... ... 0.2 ... ... ... ... 0.1 ... ... ... ... 0.43 0.03 0.23 0.1 0.3 0.13 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ Y = X ⋅W + B ࢶഋ؀ࣻ೟ ࢶഋ؀ࣻ೟ ࢶഋ؀ࣻ೟ ഛܫҗా҅ ഛܫҗా҅ ޷੸࠙೟ 0 1 0 ... 0 ⎡ ⎣ ⎤ ⎦
  13. ࣻ೟੄ ೙ਃࢿ 6 7 8 ... 5 7 5 6

    ... 3 8 4 1 ... 4 ... ... ... ... 2 0 1 5 4 3 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ W = W −α ∂L ∂W 0 0 1 ... 0 1 0 0 ... 0 0 1 0 ... 0 ... ... ... ... 0 0 0 0 0 1 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 0 1 0 ... 0 1 0 0 ... 0 0 1 0 ... 0 ... ... ... ... 0 0 0 1 0 0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 4 1 5 ... 8 ⎡ ⎣ ⎤ ⎦ 1 2 (y k − t k )2 ∑ y k = eak eai ∑ 0.5 0.4 ... ... 0.8 0.1 −0.3 ... ... 0.2 ... ... ... ... 0.1 ... ... ... ... 0.43 0.03 0.23 0.1 0.3 0.13 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ Y = X ⋅W + B ഛܫҗా҅ ޷੸࠙೟ ࢶഋ؀ࣻ೟ ؘ੉ఠ಴അ ҅࢑੄ബਯࢿ ୭੸ച ౠࢿ୶୹ ഛܫ࠙ನ ୶ۿ߂৘ஏ оࢸѨૐ ೧ࢳ೟੸੽Ӕ ӝ਎ӝ҅࢑ ಞ޷࠙ ੿ӏച 0 1 0 ... 0 ⎡ ⎣ ⎤ ⎦
  14. ࢶഋ؀ࣻ೟੉ۆ? ߭ఠҕр x 11 x 12 x 13 x 14

    x 15 x 21 x 22 x 23 x 24 x 25 x 31 x 32 x 33 x 34 x 35 x 41 x 42 x 43 x 44 x 45 x 51 x 52 x 53 x 54 x 55 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ T y 1 y 2 y 3 y 4 y 5 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ −3 −2 6 ⎡ ⎣ ⎤ ⎦ 2 2 8 ⎡ ⎣ ⎤ ⎦ ߭ఠো࢑ ೯۳ো࢑ ରਗ ࢶഋߑ੿ध ݠन۞׬ ঐഐച ੿ࠁѨ࢝ ੉޷૑೐۽ࣁय ؀ӏݽ୊ܻ
  15. ਋ܻח ࢶഋ؀ࣻܳ ׮ܖӝ ਤ೧ Numpyܳ ࢎਊ೤פ׮   ৈӝח౵੉௑੉૑݅ ࣽࣻ1ZUIPO਷ખוܿ

     /VNQZח௏যо$'PSUSBOӝ߈੉ۄࡅܴ  ࡅܳࡺ݅ইפۄߓৌਸബਯ੸ਵ۽୊ܻ೧ݫݽܻب؏ࢎਊೣ  пઙಞܻೠҊࣻળੋఠಕ੉झ৬بҳٜਸઁҕ NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ
  16. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ߭ఠ (vector) ↟ ߭ఠ ҕрীࢲ੄ਗࣗܳ಴അ ↟߭ఠחпਗࣗ੄ؘ੉ఠఋੑ੉زੌ ↟ࠁాBSSBZ۽಴അ

    OVNQZBSSBZ
  17. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ߭ఠ (vector) x = x 1 x

    2 ... x n ⎡ ⎣ ⎤ ⎦ x = x 1 x 2 ... x n ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ೯߭ఠ SPXWFDUPS ৌ߭ఠ DPMVNOWFDUPS
  18. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ߭ఠ (vector)

  19. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ߭ఠ ো࢑ ↟ؔࣅ BEEJUJPO  ↟ࡓࣅ TVCUSBDUJPO

     ↟झணۄғ 4DBMBS1SPEVDU  ↟ࢿ࠙ғ &MFNFOUXJTF.VMUJQMJDBUJPO  ↟ղ੸ *OOFS1SPEVDU ߭ఠח׮਺੄ো࢑ٜਸࣻ೯ೡࣻ੓਺
  20. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ߭ఠ ো࢑ 1 ↟ؔࣅ BEEJUJPO  ↟ࡓࣅ

    TVCUSBDUJPO v + u = (v 1 + u 1 ,...,v n + u n ) v − u = (v 1 − u 1 ,...,v n − u n )
  21. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ߭ఠ ো࢑ 1

  22. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ߭ఠ ো࢑ 2 ↟झணۄғ 4DBMBS1SPEVDU  ↟ࢿ࠙ғ

    &MFNFOUXJTF.VMUJQMJDBUJPO  ↟ղ੸ *OOFS1SPEVDU v⊗u = (v 1 u 1 ,...,v n u n ) av = (av 1 ,...,av n ) v⋅u = v i u i i=1 n ∑
  23. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ߭ఠ ো࢑ 2

  24. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ೯۳ (Matrix) ↟NYO੄ഋకܳо૗ ↟п೯ৌਸ߭ఠ۽಴അоמ ↟ࠁాରਗBSSBZ۽಴അ

  25. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ೯۳ ো࢑ ↟ؔࣅ BEEJUJPO  ↟ࡓࣅ TVCUSBDUJPO

     ↟झணۄғ 4DBMBS1SPEVDU  ↟ࢿ࠙ғ &MFNFOUXJTF.VMUJQMJDBUJPO  ↟೯۳ғ .BUSJY.VMUJQMJDBUJPO ೯۳਷׮਺੄ো࢑ٜਸࣻ೯ೡࣻ੓਺
  26. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ೯۳ ো࢑ 1 ↟ؔࣅ BEEJUJPO  ↟ࡓࣅ

    TVCUSBDUJPO X +Y = x 11 + y 11 ... x 1n + y 1n ... ... ... x m1 + y m1 ... x mn + y mn ⎡ ⎣ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ X −Y = x 11 − y 11 ... x 1n − y 1n ... ... ... x m1 − y m1 ... x mn − y mn ⎡ ⎣ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥
  27. ೯۳ ো࢑ 1 NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ

  28. ೯۳ ো࢑ 2 ↟झணۄғ 4DBMBS1SPEVDU  ↟ࢿ࠙ғ &MFNFOUXJTF.VMUJQMJDBUJPO aX =

    ax 11 ... ax 1n ... ... ... ax m1 ... ax mn ⎡ ⎣ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ X ⊗Y = x 11 y 11 ... x 1n y 1n ... ... ... x m1 y m1 ... x mn y mn ⎡ ⎣ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ
  29. ೯۳ ো࢑ 2 NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ

  30. ೯۳ ো࢑ 3 x 11 x 12 x 13 x

    14 x 21 x 22 x 23 x 24 x 31 x 32 x 33 x 34 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ೯۳ғ .BUSJY.VMUJQMJDBUJPO = y 11 y 12 y 13 y 21 y 22 y 23 y 31 y 32 y 33 y 41 y 42 y 43 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ x 11 y 11 + x 12 y 21 + x 13 y 31 + x 14 y 41 ... ... ... ... ... ... ... ... ⎡ ⎣ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ Y Y Y = Y NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ
  31. ೯۳ ো࢑ 3 NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ

  32. ੹஖ (Transpose) NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ XT = x 11 x

    12 x 13 x 14 x 21 x 22 x 23 x 24 x 31 x 32 x 33 x 34 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ T = x 11 x 12 x 13 x 21 x 22 x 23 x 31 x 32 x 33 x 41 x 42 x 43 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ xT = x 1 x 2 ... x n ⎡ ⎣ ⎤ ⎦ T = x 1 x 2 ... x n ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥
  33. ੹஖ (Transpose) NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ

  34. ױਤ ೯۳ (Identity Matrix) NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ I n =

    1 0 0 ... 0 0 1 0 ... 0 0 0 1 ... 0 ... ... ... ... 0 0 0 0 0 1 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥
  35. NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ױਤ ೯۳ (Identity Matrix)

  36. ৉೯۳ (Inverse Matrix) XX−1 = I n = X−1X NumPy۽

    ࢶഋ؀ࣻ೟ ׮ܞࠁӝ ৉೯۳਷ ೦࢚ ઓ੤ೞ૑ח ঋ਺
  37. ৉೯۳ (Inverse Matrix) NumPy۽ ࢶഋ؀ࣻ೟ ׮ܞࠁӝ

  38. ୶о ӝୡ ࣻ೟ ࢓ಝࠁӝ ੿ӏച 1 y i = x

    i x i ∑ ੹୓ sumਵ۽ ա׃ਵ۽ॄ [0, 1] ҳрਵ۽ ੿ӏച
  39. ୶о ӝୡ ࣻ೟ ࢓ಝࠁӝ ੿ӏച 1

  40. ӝఋ ӝୡ ࣻ೟ ࢓ಝࠁӝ ੿ӏച 2 Xi = X i

    − E(X) σ E(X) = 1 N X i i=1 N ∑ σ = 1 N (X i − E(X))2 i=1 N ∑ ಴ળ ੿ӏ ࠙ನܳ ഝਊ೧ ಣӐ = 0, ࠙࢑ = 1۽ ੿ӏച
  41. ӝఋ ӝୡ ࣻ೟ ࢓ಝࠁӝ ੿ӏച 2

  42. ӝఋ ӝୡ ࣻ೟ ࢓ಝࠁӝ ޷࠙ ߂ Ӓۄ٣঱౟ 1 2 3

    0 0 1 ... 0 1 0 0 ... 0 0 1 0 ... 0 ... ... ... ... 0 0 0 0 0 1 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 0 1 0 ... 0 1 0 0 ... 0 0 1 0 ... 0 ... ... ... ... 0 0 0 1 0 0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ৘ஏ ݽ؛ীࢲ੄ Ѿҗчҗ पઁ Ѿҗчҗ੄ рӓ੄ ૑಴ੋ ର੉ “ࣚप"ਸ ઴੉ӝ ਤೣ ࣚप ೣࣻ Ӓې೐
  43. ӝఋ ӝୡ ࣻ೟ ࢓ಝࠁӝ ಞ޷࠙ f (x) = g(x 1

    )+...+ g(x n ) ∂ f (x) ∂x 1 = ∂g(x 1 ) ∂x 1 ׮߸ࣻ ҕр ഑਷ ೣࣻীࢲ ౠ੿ ߸ࣻܳ ؀࢚ਵ۽ ޷࠙ ݠन ۞׬ীࢶ ࠁా ೯۳ਸ ؀࢚ਵ۽ೠ ೯۳ ಞ޷࠙ ࢎਊ ∂L ∂X = ∂L ∂x 11 ... ∂L ∂x 1n ... ... ... ∂L ∂x m1 ... ∂L ∂x mn ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥
  44. ࢶഋ ഥӈ ҳഅ೧ࠁӝ ਤ੄ѐ֛ٜਸഝਊ೧рױೠ ࢶഋഥӈݽ؛ਸٜ݅যࠁѷणפ׮

  45. ࢶഋ ഥӈ ҳഅ೧ࠁӝ ୭੸੄ ࢶഋ ࢚ҙ ҙ҅ܳ ಴അೞח ૒ࢶਸ ଺ਵ۰Ҋ

    ೣ → ୭੸੄ ߬ఋܳ ೟ण y = β 0 x 0 + β 1 x 1 y i = x i T β
  46. ࢶഋ ഥӈ ҳഅ೧ࠁӝ y = β 0 x 0 +

    β 1 x 1 ױੌؘ੉ఠ
  47. ࢶഋ ഥӈ ҳഅ೧ࠁӝ y = β 0 x 0 +

    β 1 x 1 y i = x i0 β 0 + x i1 β 1 ױੌؘ੉Tఠ ׮઺ؘ੉ఠ઺Jߣ૩ؘ੉ఠ
  48. ࢶഋ ഥӈ ҳഅ೧ࠁӝ y = β 0 x 0 +

    β 1 x 1 y i = x i T β y i = x i0 β 0 + x i1 β 1 ױੌؘ੉ఠ ׮઺ؘ੉ఠ઺Jߣ૩ؘ੉ఠ Jߣ૩ؘ੉ఠ߭ఠ಴അ
  49. ࢶഋ ഥӈ ҳഅ೧ࠁӝ Y = Xβ Y = y 1

    y 2 ... y n ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ X = x 1 T x 2 T ... x n T ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ β = β 0 β 1 ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ੌ߈ചػ׮઺ؘ੉ఠ੄೯۳಴അ
  50. ࢶഋ ഥӈ ҳഅ೧ࠁӝ 1.0 1.1 1.2 ... 10.0 ⎡ ⎣

    ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥
  51. ࢶഋ ഥӈ ҳഅ೧ࠁӝ 1.0 1.1 1.2 ... 10.0 ⎡ ⎣

    ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 0.1 0.11 0.12 ... 1.0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥
  52. ࢶഋ ഥӈ ҳഅ೧ࠁӝ y = ax + b = b

    + ax y 1 = b + ax 1 y 2 = b + ax 2
  53. ࢶഋ ഥӈ ҳഅ೧ࠁӝ 1 0.1 1 0.11 1 0.12 1

    ... 1 1.0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥
  54. ࢶഋ ഥӈ ҳഅ೧ࠁӝ r 1 r 2 r 3 ...

    r 100 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 1 0.1 1 0.11 1 0.12 1 ... 1 1.0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥
  55. ࢶഋ ഥӈ ҳഅ೧ࠁӝ r 1 r 2 r 3 ...

    r 100 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 1 x i1 1 x i2 1 x i3 ... ... 1 x i20 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ 1 0.1 1 0.11 1 0.12 1 ... 1 1.0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥
  56. ࢶഋ ഥӈ ҳഅ೧ࠁӝ Ӓۄ٣঱౟ ࢸ҅ 1 0.1 1 0.11 1

    0.12 ... ... 1 1.0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ w 0 w 1 ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥
  57. ࢶഋ ഥӈ ҳഅ೧ࠁӝ Ӓۄ٣঱౟ ࢸ҅ 1 0.1 1 0.11 1

    0.12 ... ... 1 1.0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ w 0 w 1 ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ y 1 y 2 y 3 ... y 20 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ − w 0 + 0.1w 1 w 0 + 0.11w 1 w 0 + 0.12w 1 ... w 0 +1.0w 1 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎛ ⎝ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎞ ⎠ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ T
  58. ࢶഋ ഥӈ ҳഅ೧ࠁӝ Ӓۄ٣঱౟ ࢸ҅ 1 0.1 1 0.11 1

    0.12 ... ... 1 1.0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ w 0 w 1 ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ y 1 y 2 y 3 ... y 20 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ − w 0 + 0.1w 1 w 0 + 0.11w 1 w 0 + 0.12w 1 ... w 0 +1.0w 1 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎛ ⎝ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎞ ⎠ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ T 1 20 error i 2 i=1 20 ∑ = 1 20 (y i − e i )2 i=1 20 ∑
  59. ࢶഋ ഥӈ ҳഅ೧ࠁӝ Ӓۄ٣঱౟ ࢸ҅ 1 0.1 1 0.11 1

    0.12 ... ... 1 1.0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ w 0 w 1 ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ y 1 y 2 y 3 ... y 20 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ − w 0 + 0.1w 1 w 0 + 0.11w 1 w 0 + 0.12w 1 ... w 0 +1.0w 1 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎛ ⎝ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎞ ⎠ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ T 1 20 error i 2 i=1 20 ∑ = 1 20 (y i − e i )2 i=1 20 ∑ ∂(mse) ∂W = [error][x]
  60. ࢶഋ ഥӈ ҳഅ೧ࠁӝ Ӓۄ٣঱౟ ࢸ҅ ∂(mse) ∂W ೯۳ಞ޷࠙→ ೯۳੄ п

    ਗࣗী ؀೧ ಞ޷࠙
  61. ࢶഋ ഥӈ ҳഅ೧ࠁӝ Ӓۄ٣঱౟ ࢸ҅ ∂(mse) ∂w 0 = ∂(

    (y i − (w 0 + x i w 1 ))2 ) ∑ ∂w 0 = (y i − (w 0 + x i w 1 ))⋅1 ∑ ∂(mse) ∂w 1 = ∂( (y i − (w 0 + x i w 1 ))2 ) ∑ ∂w 1 = (y i − (w 0 + x i w 1 ))⋅ x i ∑
  62. ࢶഋ ഥӈ ҳഅ೧ࠁӝ Ӓۄ٣঱౟ ࢸ҅ ∂(mse) ∂W = [ y

    i − (w 0 + x i w 1 ) ∑ ] 1 0.1 1 0.11 1 0.12 ... ... 1 1.0 ⎡ ⎣ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ = [error][x]
  63. ࢶഋ ഥӈ ҳഅ೧ࠁӝ पઁ ೟ण ױ҅ W ←W −α ∂(mse)

    ∂W
  64. ࢶഋ ഥӈ ҳഅ೧ࠁӝ ೟ण Ӓې೐ <ഥ>&SSPS <ഥ>&SSPS <ഥ>&SSPS <ഥ>&SSPS

  65. ࢶഋ ഥӈ ҳഅ೧ࠁӝ ୭ઙ ೟ण Ѿҗ &SSPS β = −2.152736

    11.075026 ⎡ ⎣ ⎢ ⎤ ⎦ ⎥ y k = x k T β →
  66. Next (more LA and Mathematics) ↟઱ࢿ࠙࠙ࢳ 1$"  ↟ױੌч࠙೧ 47%

     ↟-6࠙೧ ↟ҊਬчҊਬ߭ఠ ↟಴ળച ↟j ↟߬੉૑উా҅ ↟ഛܫӏ஗ ↟ࢠ೒݂ߑध ↟୭؀਋بஏ੿ ↟j ↟೯۳޷੸࠙೟ ↟Ӓۄ٣঱౟ ↟j
  67. хࢎ೤פ׮ MinJae Kwon (@mingrammer) Getting started to learn the linear

    algebra with python 2017.08.12