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

PRML(分類編)

Sponsored · Your Podcast. Everywhere. Effortlessly. Share. Educate. Inspire. Entertain. You do you. We'll handle the rest.
Avatar for gucchi gucchi
September 06, 2019

 PRML(分類編)

Avatar for gucchi

gucchi

September 06, 2019
Tweet

More Decks by gucchi

Other Decks in Science

Transcript

  1. ໨࣍ 1. ༧උ஌ࣝ 1-1. ෼ྨ໰୊ͷΞϓϩʔν 2. ϑΟογϟʔͷઢܗ൑ผ (PRML 4.1.4) 3.

    ϩδεςΟοΫճؼ (PRML 4.3, 4.4, 4.5) 3-1. ϩδεςΟοΫճؼ 3-2. ϩδεςΟοΫճؼͷ࠷໬ਪఆ 3-3. ϕΠζϩδεςΟοΫճؼ 3 / 47
  2. 1. ༧උ஌ࣝ ػցֶशɺಛʹͦͷதͰ΋ڭࢣ͋ΓֶशͰ͸ɺ·ͣೖྗσʔλͷू߹ {x1 , x2 , · · ·

    , xN } ͱͦΕͧΕʹରԠ͢Δ໨ඪϕΫτϧͷू߹ {t1 , t2 , · · · , tN } Λ༻ҙ͢Δɻ(܇࿅σʔλɺ·ͨ͸ڭࢣσʔλ) ༻ҙͨ͠܇࿅σʔλΛ༻͍ͯɺೖྗσʔλ͔Β໨ඪϕΫτϧΛ༧ଌ͢Δ ؔ਺ y(x) Λ࡞Δɻ(ֶश) ֶशऴྃޙɺະ஌ͷσʔλ x ͷ໨ඪϕΫτϧΛ y(x) Ͱ༧ଌ͢Δ ֤ೖྗϕΫτϧΛ༗ݶݸͷ཭ࢄతͳΧςΰϦʹׂΓ౰ͯΔ৔߹ (ྫ͑ ͹ɺखॻ͖਺ࣈͷೝࣝ) ΛΫϥε෼ྨͱ͍͍ɺग़ྗ͕࿈ଓม਺ͷ৔߹ Λճؼͱ͍͏ɻ ࠓճ͸Ϋϥε෼ྨͷ໰୊Λߟ͑Δɻ 4 / 47
  3. 1-1. ෼ྨ໰୊ͷΞϓϩʔν ·ͣ͸ɺ܇࿅σʔλ͕༩͑ΒΕͨ࣌ͷ෼ྨ໰୊ͷΞϓϩʔνΛ 2 ͭ঺հ ͢Δɻ 1 ͭ໨͸ɺ܇࿅σʔλ͔Βࣝผؔ਺ y(x) Λ࡞Δํ๏Ͱ͋Δɻ

    ͜͜Ͱ x ͸ D ࣍ݩͷೖྗϕΫτϧ (ྫ͑͹ɺn ݸ໨ͷ܇࿅σʔλͷೖྗ Ͱ͋Δ xn ) Ͱ͋Δɻ ͜ͷΞϓϩʔνͰ͸ɺy(x) ͷ஋Λ༻͍ͯɺ܇࿅σʔλʹͳ͍ະ஌ͷ σʔλ x ͕ͲͷΫϥεʹ෼ྨ͞ΕΔ΂͖ͳͷ͔Λ༧ଌ͢Δɻ(ޙͰྫΛ ͋͛Δ͕ɺྫ͑͹ x ͸ y(x) ≥ 0 ͳΒΫϥε 1, y(x) < 0 ͳΒΫϥε 2 ʹଐ͢ΔͳͲ) 5 / 47
  4. 1-1. ෼ྨ໰୊ͷΞϓϩʔν ࣮ͨͩ͠ࡍʹ෼ྨ໰୊Λղ͘ͱ͖͸ɺ܇࿅σʔλΛ࢖ͬͯؔ਺ y(x) Λ Ұ͔Β࡞Γ্͛Δ͜ͱ͸͠ͳ͍ɻ Α͘ߦ͏ํ๏͸ɺೖྗϕΫτϧͱಉ࣍͡ݩͷύϥϝʔλϕΫτϧ (ॏ Έ)w =

    (w1 , w2 , · · · , wD )T ͱεΧϥʔύϥϝʔλ (όΠΞε)w0 Λ༻ҙ ͠ɺؔ਺ y(x, w, w0 ) y(x, w, w0 ) = f ( wTx + w0 ) (1.1) Λ༻ҙ͢Δɻ ͜͜Ͱɺf ͸ඇઢܗؔ਺Ͱ͋Γɺ׆ੑԽؔ਺ͱݺ͹ΕΔɻ(ྫ͑͹ϩδ εςΟοΫγάϞΠυؔ਺) ͦͯ͠ɺڭࢣσʔλʹΑΓؔ਺ y(x, w, w0 ) ͷύϥϝʔλ w, w0 Λ఺ਪ ఆ͠ɺਪఆ͞Εͨύϥϝʔλͷ஋Λ w⋆, w⋆ 0 ͱ͢Δͱɺؔ਺ y(x, w = w⋆, w0 = w⋆ 0 ) ͕ࣝผؔ਺Ͱ͋Δɻ(ύϥϝʔλͷௐ੔ํ๏͸ ͨ͘͞Μ͋Δɻ) ͜Ε͕ࣝผؔ਺Λ࡞੒͢Δํ๏Ͱ͋Γɺࠓճ͸ͦͷதͰ΋ϑΟογϟʔ ͷઢܗ൑ผͱ͍͏ํ๏Λ঺հ͢Δɻ 6 / 47
  5. 1-1. ෼ྨ໰୊ͷΞϓϩʔν ΋͏ҰͭͷΞϓϩʔνͱͯ͠ɺࣝผؔ਺ y(x) ͷ࡞੒Ͱͳ͘ɺ৚݅෇͖ ֬཰ p(Ck |x) Λ܇࿅σʔλ͔Βܾఆ͢Δํ๏͕͋Δɻ ͜͜Ͱɺ෼ྨΫϥε͸

    K ݸ (Ϋϥε 1, Ϋϥε 2, · · · , Ϋϥε K) ͋Δ ͱͯ͠ɺCk ͸ k ݸ໨ͷΫϥεΛද͢ɻ ͳͷͰɺ৚݅෇͖֬཰ p(Ck |x) ͸ɺԿ͔͋Δೖྗ x ͕༩͑ΒΕͨͱ͖ʹ ͦͷೖྗ͕ k ݸ໨ͷΫϥεʹଐ͢Δ֬཰Λ༩͑Δɻ ͜Ε΋࣮ࡍ͸ɺ܇࿅σʔλ͔Β p(Ck |x) ͷܗΛҰ͔ΒܾΊΔͷͰ͸ͳ ͘ɺܗ͸ܾΊ͓͍ͯͯύϥϝʔλΛ܇࿅σʔλ͔ΒܾΊΔɻ(࠷໬ਪఆ Λ͢Δɻ) 7 / 47
  6. 1-1. ෼ྨ໰୊ͷΞϓϩʔν ͨͩ͠৚݅෇͖֬཰ͷࣝผؔ਺ͱҟͳΔͱ͜Ζ͸ɺ৚݅෇͖֬཰ͱͯ͠ Ծఆ͢Δؔ਺ͷܗͱͯ͠ɺن֨Խ৚݅ ∑ k p(Ck |x) = 1

    (1.2) Λຬͨ͢Α͏ͳؔ਺ΛԾఆ͠ͳͯ͘͸͍͚ͳ͍ɻ ͜ͷΑ͏ʹ৚݅෇͖֬཰ΛϞσϧԽ͢Δ͜ͱʹΑͬͯɺ࠷໬ਪఆ͔Βϕ ΠζਪఆʹࣗવʹҠߦͰ͖Δɻ ࠓճ͸۩ମྫͱͯ͠ɺK = 2 ͷ࣌ͷϩδεςΟοΫճؼͷ࠷໬ਪఆͱ ϕΠζਪఆΛऔΓѻ͏ɻ 8 / 47
  7. 2. ϑΟογϟʔͷઢܗ൑ผ ͦΕͰ͸ɺࣝผؔ਺Λ܇࿅σʔλ͔ΒܾΊΔํ๏ͷ۩ମྫͱͯ͠ϑΟο γϟʔͷઢܗ൑ผΛઆ໌͢Δɻ ·ͣ͸ 2 Ϋϥε෼ྨΛߟ͑Δɻ(ΫϥεͷϥϕϧΛ C1 ͱ C2

    ͱ͢Δɻ) ࣝผؔ਺ͱͯ͠ɺ(1.1) ͷ׆ੑؔ਺ f Λ߃౳ؔ਺ͱͨ͠΋ͷΛߟ͑Δɻ y(x, w, w0 ) = wTx + w0 (2.1) ͦͯ͠ɺN ݸͷ܇࿅σʔλͷೖྗσʔλͱͯ͠ {x1 , x2 , · · · , xN } ͱ͠ɺ ͦΕΒʹରԠ͢Δ໨ඪม਺ͱͯ͠ {t1 , t2 , · · · , tN } ͱ͠ɺศ্ٓΫϥε C1 ͕ t = 1 ʹରԠ͠ɺΫϥε C2 ͕ t = 0 ʹରԠ͢Δͱ͢Δɻ ·ͨɺೖྗϕΫτϧ͕ॴଐ͢ΔΫϥεͷ൑ఆํ๏Ͱ͋Δ͕ɺ͋Δೖྗϕ Ϋτϧ xi ͕༩͑ΒΕͨ࣌ʹ y(xi , w, w0 ) ≥ 0 Ͱ͋Ε͹ xi ͸Ϋϥε C1 ʹॴଐ͠ (ͭ·Γ ti = 1)ɺy(xi , w, w0 ) < 0 Ͱ͋Ε͹ xi ͸Ϋϥε C2 ʹ ॴଐ͢Δ (ͭ·Γ ti = 0) ͱ͢Δɻ 9 / 47
  8. 2. ϑΟογϟʔͷઢܗ൑ผ ͦ΋ͦ΋ɺD ࣍ݩͷೖྗσʔλ x Λ 1 ࣍ݩͷؔ਺ y(x, w,

    w0 ) ʹࣹӨ͠ ͯ y ͷਖ਼ෛͰॴଐ͢ΔΫϥεΛܾΊΔͷͰɺ࣍ݩͷݮগʹΑΓ൑அ͢Δ ͨΊͷ৘ใ͕େ͖͘ݮͬͨྔ (ͭ·Γ y) ͰΫϥεΛ൑அ͢Δ͜ͱʹ ͳΔɻ ͭ·ΓɺD ۭؒͰ͸ೖྗσʔλ͕ଐ͢ΔΫϥεͷྖҬಉ͕࢜Α͘෼཭ ͞Ε͍͕ͯͨɺ1 ࣍ݩ΁ͷࣹӨͷ࢓ํʹΑͬͯ͸ɺͦͷ෼཭͕ͳ͘ͳΔ (ॏͳͬͯ͠·͏) ͜ͱ͕͋Δɻ D = 2 ͷ࣌ͷ۩ମྫΛ࣍ͷεϥΠυͰࣔ͢ɻ 10 / 47
  9. 2. ϑΟογϟʔͷઢܗ൑ผ ੨ͷϓϩοτ͕Ϋϥε C1 ʹॴଐ͢Δ܇࿅σʔλɺ੺ͷϓϩοτ͕Ϋϥ ε C2 ʹॴଐ͢Δ܇࿅σʔλͱ͠ɺը૾ 2 ຕͱ΋܇࿅σʔλ͸શ͘ಉ͡

    Ͱ͋Δɻ y(x, w, w0 ) = 0 ͷ௚ઢ͸ύϥϝʔλ w, w0 ΛมԽͤ͞Δ͜ͱʹΑͬͯɺ D = 2 ͷೖྗۭؒΛॎԣແਚʹҠಈ͢Δɻ ੨ͷϓϩοτ͕ଟ͍ํΛ y(x, w, w0 ) ≥ 0 ͷྖҬͱͯ͠ɺ੺ͷϓϩοτ ͕ଟ͍ํΛ y(x, w, w0 ) < 0 ͷྖҬͱ͢Δɻ ྆ํͷΫϥϑͰڭࢣσʔλ͸౳͘͠ɺೖྗۭؒͰ͸Α͘෼཭͞Ε͍ͯΔ ͕ɺࠨଆͷࣹӨ͸ͦͷ෼཭ΛҰ෦ফͯ͠͠·͍ͬͯΔ͜ͱ͕Θ͔Δɻ 11 / 47
  10. 2. ϑΟογϟʔͷઢܗ൑ผ ϑΟογϟʔͷઢܗ൑ผͰ͸ɺ܇࿅σʔλΛ࠷΋෼཭͢ΔΑ͏ͳࣹӨɺ ͭ·Γύϥϝʔλ w, w0 ΛٻΊΔ͜ͱΛߟ͑Δɻ ͦ͜Ͱ·ͣɺ܇࿅σʔλͷೖྗ {x1 ,

    x2 , · · · , xN } ͷΫϥεผͷฏۉϕ ΫτϧΛ m1 , m2 ͱ͠ɺ m1 = 1 N1 ∑ n∈C1 xn , m2 = 1 N2 ∑ n∈C2 xn (2.2) ͱͳΔɻ ͜͜ͰɺN1 ͱ N2 ͸ͦΕͧΕɺΫϥε C1 ·ͨ͸ C2 ʹଐ͍ͯ͠Δ܇࿅ σʔλͷ਺Ͱ͋Δɻ(΋ͪΖΜɺN1 + N2 = N Λຬͨ͢ɻ) 12 / 47
  11. 2. ϑΟογϟʔͷઢܗ൑ผ ҰํɺࣹӨޙͷΫϥεผͷฏۉ͸ y(x, w, w0 )(2.1) ͷઢܗੑΑΓɺҎԼ ͷΑ͏ʹͳΔɻ m1

    = 1 N1 ∑ n∈C1 y(xn , w, w0 ) = wTm1 + w0 m2 = 1 N2 ∑ n∈C2 y(xn , w, w0 ) = wTm2 + w0 (2.3) Α͘෼཭͞ΕࣹͨӨͰ͸ɺ͜ΕΒͷࣹӨޙͷΫϥεผͷฏۉ͸େ͖͘ҟ ͳ͍ͬͯΔͱߟ͑ΒΕΔɻ ͭ·Γɺ m2 − m1 = wT(m2 − m1 ) (2.4) ͷ஋͕େ͖͍΄ͲɺࣹӨޙ΋ྑ͘෼཭͞Ε͍ͯΔͩΖ͏ɻ 13 / 47
  12. 2. ϑΟογϟʔͷઢܗ൑ผ ͨͩ͠ɺࣹӨޙͷΫϥεؒͷฏۉ஋ͷ͕ࠩେ͖ͯ͘΋ɺࣹӨޙͷΫϥε ผͷ෼ࢄ͕େ͖͔ͬͨΒɺ̍࣍ݩʹࣹӨ͞ΕͨΫϥεͷྖҬ͕ॏͳΓ ߹ͬͯ͠·͏ɻ ͦ͜ͰɺҎԼͷΫϥεผͷ෼ࢄͷ஋͸খ͘͞ͳ͍ͬͯͯ΄͍͠ɻ s2 1 = ∑

    n∈C1 {y(xn , w, w0 ) − m1 }2 s2 2 = ∑ n∈C2 {y(xn , w, w0 ) − m2 }2 (2.5) ΑͬͯɺҎԼͷϑΟογϟʔͷ൑ผج४Λ࠷େʹ͢ΔΑ͏ͳύϥϝʔλ w, w0 ͕ࣹӨޙ΋෼཭Λ࠷େʹอͭΑ͏ͳύϥϝʔλͰ͋Δ͜ͱ͕Θ ͔Δɻ J(w, w0 ) = (m2 − m1 )2 s2 1 + s2 2 (2.6) 14 / 47
  13. 2. ϑΟογϟʔͷઢܗ൑ผ ͦΕͰ͸ɺ(2.6) ͷӈลΛ w, w0 Ͱॻ͖Լͦ͏ɻ ·ͣ෼ࢠ͸ (2.4) ΑΓɺ

    (m2 − m1 )2 = { wT(m2 − m1 ) }2 = wT(m2 − m1 )(m2 − m1 )Tw = wTSB w (2.7) ͱͳΔɻ ͜͜ͰɺSB ͸Ϋϥεؒڞ෼ࢄߦྻͱ͍͍ɺ SB = (m2 − m1 )(m2 − m1 )T (2.8) Ͱఆٛ͞ΕΔɻ 15 / 47
  14. 2. ϑΟογϟʔͷઢܗ൑ผ ࣍ʹɺ(2.6) ͷ෼฼͸ s2 1 + s2 2 Ͱ͋Δ͕ɺͱΓ͋͑ͣ

    s2 1 ͚ͩ w, w0 Ͱॻ ͍ͯΈΔͱɺ(2.5) ΑΓɺ s2 1 = ∑ n∈C1 {wTxn + w0 − m1 }2 = ∑ n∈C1 {wTxn − wTm1 }2 = ∑ n∈C1 {wT(xn − m1 )}2 = ∑ n∈C1 wT(xn − m1 )(xn − m1 )Tw =wT [ ∑ n∈C1 (xn − m1 )(xn − m1 )T ] w (2.9) ͱͳΔɻ 16 / 47
  15. 2. ϑΟογϟʔͷઢܗ൑ผ ಉ༷ͷมܗΛ s2 2 ʹ͍ͭͯ΋ߦ͏ͱɺ(2.6) ͷ෼฼ s2 1 +

    s2 2 ͸ҎԼͷΑ͏ ʹͳΔɻ s2 1 + s2 2 = wTSW w (2.10) ͜͜ͰɺSW ͸૯Ϋϥε಺ڞ෼ࢄߦྻͱ͍͍ɺҎԼͰఆٛ͞ΕΔɻ SW = ∑ n∈C1 (xn − m1 )(xn − m1 )T + ∑ n∈C2 (xn − m2 )(xn − m2 )T (2.11) ͜ΕΑΓɺϑΟογϟʔͷ൑ผج४ (2.6) ͸ w Λ༻͍ͯҎԼͷΑ͏ʹॻ ͚Δɻ J(w, w0 ) = J(w) = wTSB w wTSW w (2.12) ͜͜Ͱɺ݁ՌతʹϑΟογϟʔͷ൑ผج४͸ w ʹͷΈґଘ͠ɺw0 ʹ͸ ґଘ͠ͳ͍͜ͱ͕Θ͔Δɻ 17 / 47
  16. 2. ϑΟογϟʔͷઢܗ൑ผ ͜Ε͔ΒϑΟογϟʔͷ൑ผج४ J(w) Λ࠷େʹ͢Δ w ΛٻΊ͍ͯ͘ ͜ͱʹͳΔ͕ɺ͜͜Ͱ 1 ͭ

    J(w) ʹ͸ॏཁͳੑ࣭͕͋Δɻ θϩͰ͸ͳ͍ఆ਺ α Λ༻ҙ͠ɺJ(αw) Λܭࢉ͢Δͱɺ J(αw) = (αw)TSB (αw) (αw)TSW (αw) = wTSB w wTSW w = J(w) (2.13) ͱͳΓɺϑΟογϟʔͷ൑ผج४ J(w) ͸ w ͷఆ਺ഒʹରͯ͠ෆมͰ ͋Δɻ(εέʔϧෆมੑ) ͜ͷੑ࣭ΑΓɺJ(w) Λ࠷େʹ͢Δ w(= w⋆) ͕ݟ͔ͭΔͱɺͦͷఆ਺ ഒͷ αw⋆ ΋ J(w) Λ࠷େʹ͢Δͱ͍͏͜ͱ͕Θ͔Δɻ ͜ΕΑΓɺզʑ͸ J(w) Λ࠷େʹ͢Δ w ΛٻΊΔࡍɺ࠷େʹ͢Δ w ͸ Ͳͷํ޲Λ޲͍͍ͯΔ͔͚ͩΛ஌Ε͹ྑ͍ࣄʹͳΔɻ 18 / 47
  17. 2. ϑΟογϟʔͷઢܗ൑ผ ͦΕͰ͸࣮ࡍʹ J(w) Λ࠷େʹ͢Δ w ΛٻΊΔҝʹɺJ(w) Λ w Ͱඍ

    ෼͢Δͱɺ ∂J(w) ∂w = 2 (wTSW w)2 { SB w(wTSW w) − SW w(wTSB w) } (2.14) ͱͳΓɺ͜ΕΛθϩʹ͢Δ w ͸ҎԼͷࣜΛຬͨ͢ɻ SB w(wTSW w) − SW w(wTSB w) = 0 (2.15) ྆ลɺࠨ͔Β S−1 W Λ͔͚ͯɺগࣜ͠มܗΛ͢Δͱɺ S−1 W SB w(wTSW w) − w(wTSB w) = 0 →w = (wTSW w) (wTSB w) S−1 W SB w w ∝ S−1 W SB w (2.16) ͱͳΔ͜ͱ͕Θ͔Δɻ 19 / 47
  18. 2. ϑΟογϟʔͷઢܗ൑ผ ͜͜ͰɺSB w ͸ SB w = (m2 −

    m1 )(m2 − m1 )Tw ∝ m2 − m1 (2.17) ͱͳΔ͜ͱΛ༻͍ΔͱɺϑΟογϟʔͷ൑ผج४ J(w) Λ࠷େʹ͢Δϕ Ϋτϧ w ͸ w ∝ S−1 W (m2 − m1 ) (2.18) ͱͳΓɺS−1 W (m2 − m1 ) ͱಉ͡ํ޲Λ޲͍͍ͯΔϕΫτϧͰ͋Δ͜ͱ͕ Θ͔Δɻ Αͬͯɺߴ࣍ݩͷೖྗσʔλΛ (2.1) Λ࢖ͬͯ̍࣍ݩʹࣹӨ͠ɺΫϥε Λ൑அ͢ΔࡍɺϑΟογϟʔͷઢܗ൑ผΛ༻͍ΔͱɺS−1 W (m2 − m1 ) ํ ޲ͷύϥϝʔλ w Λ༻͍ࣹͯӨΛߦ͏ͱɺೖྗσʔλͷ෼཭Λ࠷େݶ ʹอͪͳ͕ΒࣹӨ͢Δ͜ͱ͕Ͱ͖Δ͜ͱ͕Θ͔ͬͨɻ 20 / 47
  19. 3. ϩδεςΟοΫճؼ ͜Ε·Ͱ͸ࣝผؔ਺ y(x) Λڭࢣσʔλ͔Βܾఆ͢Δํ๏ΛऔΓѻͬͨ ͕ɺ͔͜͜Β͸৚݅෇͖֬཰ p(Ck |x) Λܾఆ͢Δํ๏ΛऔΓѻ͏ɻ ࣝผؔ਺ͷ࣌ͱಉ͡Α͏ʹɺ܇࿅σʔλͱͯ͠ೖྗσʔλͷू߹

    {x1 , x2 , · · · , xN } ͱͦΕͧΕʹରԠ͢Δ໨ඪม਺ͷू߹ {t1 , t2 , · · · , tN } Λ༻ҙ͢Δɻ ࠓճ͸Ϋϥε਺ K ͕ 2 ͷ࣌ (ೋ஋෼ྨ) Λѻ͏ͨΊɺ໨తม਺ tn ͸ 0 ͔ 1 ͷ཭ࢄతͳ஋ΛͱΔɻ ࠓճ͸෼ྨͷϞσϧͱͯ͠ɺϩδεςΟοΫճؼϞσϧΛ঺հ͢Δɻ (ʮճؼʯͱ෇͍͍ͯΔ͕ɺ෼ྨͷϞσϧͰ͋Δɻ) 21 / 47
  20. 3-1. ϩδεςΟοΫճؼ ࣝผؔ਺ͷ࣌ͱಉ༷ʹɺ܇࿅σʔλΛ༻͍ͯ৚݅෇͖֬཰ p(Ck |x) ΛҰ ͔Β࡞Δ͜ͱ͸ͤͣɺ৚݅෇͖֬཰ p(Ck |x) ʹύϥϝʔλ

    w Λಋೖ͠ ͯɺp(Ck |x, w) Λߟ͑Δɻ ࠓճͷϩδεςΟοΫճؼͰ͸ɺΫϥε 1 ʹରͯ͠ɺҎԼͷΑ͏ͳؔ਺ p(C1 |x, w) Λߟ͑Δɻ p(C1 |x, w) = σ(wTϕ(x)) (3.1) 22 / 47
  21. 3-1. ϩδεςΟοΫճؼ ·ͣɺ(3.1) ͷӈลͷҾ਺ͷதʹ͋ΔϕΫτϧؔ਺ ϕ(·) ͸ඇઢܗͳؔ਺ ϕj (x) (j =

    0, · · · , M − 1) Λॎʹฒ΂ͨϕΫτϧؔ਺ ϕ(x) = (ϕ0 (x), ϕ1 (x), · · · , ϕM−1 (x))T Ͱ͋Δɻ ྫ͑͹ɺجఈؔ਺ ϕj (x) ͱͯ͠ҎԼͷΨ΢εجఈؔ਺͕͋Δɻ ϕj (x) = exp { − (x − µj )2 2s2 } (3.2) ͜ͷجఈؔ਺͸ x = µj Λத৺ʹͯ͠ɺ෼ࢄ s2 ʹΑͬͯࢧ഑͞ΕΔ޿͕ ΓΛ࣋ͭΨ΢εجఈؔ਺Ͱ͋Δɻ Ҏ߱͸Ұൠͷجఈؔ਺ ϕj (x) Λ༻͍ͯٞ࿦͢Δɻ 23 / 47
  22. 3-1. ϩδεςΟοΫճؼ ·ͣɺؔ਺ σ(·) ͸ϩδεςΟοΫγάϞΠυؔ਺ͱݺ͹ΕɺҎԼͰఆ ٛ͞ΕΔɻ σ(x) = 1 1

    + e−x (3.3) ਤͰॻ͘ͱҎԼͷΑ͏ʹͳΔɻ ϩδεςΟοΫγάϞΠυؔ਺͸ఆٛʹΑΓɺ0 ͔Β 1 ·Ͱͷ஋ΛͱΔ ͷͰɺ(3.1) ΑΓ p(C1 |x, w) ͸ҎԼͷΑ͏ʹ֬཰ͷຬͨ͢΂͖஋ͷൣғ ಺ʹ஋ΛͱΔɻ 0 < p(C1 |x, w) < 1 (3.4) 24 / 47
  23. 3-1. ϩδεςΟοΫճؼ ҰํɺΫϥε 2 ʹରͯ͠ɺp(C2 |x, w) ͸ҎԼͷΑ͏ʹԾఆ͢Δɻ p(C2 |x,

    w) = 1 − p(C1 |x, w) = 1 − σ(wTϕ(x)) (3.5) (3.4) ΑΓɺp(C2 |x, w) ΋ҎԼͷΑ͏ʹ֬཰ͷຬͨ͢΂͖஋ͷൣғ಺ʹ ஋ΛͱΔɻ 0 < p(C2 |x, w) < 1 (3.6) ͜ΕΒͷఆٛʹΑΓɺԾఆͨؔ͠਺ p(Ck |x, w) ͸ن֨Խ৚݅ (1.2) Λຬ ͨ͢ɻ ∑ k p(Ck |x, w) = p(C1 |x, w) + 1 − p(C1 |x, w) = 1 (3.7) t = 1 ͷΫϥεΛ C1 ͱ͠ɺt = 0 ͷΫϥεΛ C2 ͱ͍ͯ͠ΔͷͰɺ໬౓ؔ ਺ p(t|x, w) ͸ p(t|x, w) = σ(wTϕ(x))t(1 − σ(y(wTϕ(x))))1−t (3.8) ͱͳΔɻ (͜ͷΑ͏ͳ෼෍ΛϕϧψʔΠ෼෍ͱ͍͏) 25 / 47
  24. 3-2. ϩδεςΟοΫճؼͷ࠷໬ਪఆ ࣍ʹ࠷໬ਪఆΛߦ͏͜ͱΛߟ͑Δɻ ͍ͭ΋ͷΑ͏ʹɺ܇࿅σʔλͱͯ͠ೖྗσʔλͷू߹ X = {x1 , x2 ,

    · · · , xN } ͱͦΕͧΕʹରԠ͢Δ໨ඪม਺ͷू߹ {t1 , t2 , · · · , tN } Λ༻ҙ͢Δɻ ڭࢣσʔλҰͭҰ͕ͭ෼෍ (3.8) ͔Βಠཱʹੜ੒͞Ε͍ͯΔͱ͢Δͱɺ ໬౓ؔ਺ p(t|X, w) ͸ҎԼͷΑ͏ʹͳΔɻ p(t|X, w) = N ∏ n=1 ytn n (1 − yn )1−tn (3.9) ͱͳΔɻ ͜͜Ͱ yn ͸ҎԼͰఆٛ͞ΕΔɻ yn = σ(wTϕ(xn )) (3.10) 26 / 47
  25. 3-2. ϩδεςΟοΫճؼͷ࠷໬ਪఆ ໬౓ؔ਺ (3.9) Λ࠷େʹ͢Δ w ΛٻΊΔ͜ͱ͸ҎԼͷෛͷର਺໬౓Λ ࠷খʹ͢Δ w ΛٻΊΔ͜ͱͱ౳ՁͰ͋Δɻ

    E(w) = − ln p(t|X, w) = − N ∑ n=1 ln { ytn n (1 − yn )1−tn } = − N ∑ n=1 { tn ln yn + (1 − tn ) ln (1 − yn ) } (3.11) ͜Ε͸ަࠩΤϯτϩϐʔޡࠩͱݺ͹ΕΔޡࠩؔ਺Ͱɺ෼ྨ໰୊ͰΑ͘࢖ ΘΕΔޡࠩؔ਺Ͱ͋Δɻ ෼ྨ໰୊ʹ͓͚ΔަࠩΤϯτϩϐʔޡࠩͷ࠷খԽ͸ɺ֬཰࿦Λ༻͍Δͱ ໬౓ؔ਺ΛϕϧψʔΠ෼෍ (3.8) ͱԾఆͨ͠ͱ͖ͷ࠷໬ਪఆͷ݁ՌͰ͋ Δࣄ͕Θ͔Δɻ 27 / 47
  26. 3-2. ϩδεςΟοΫճؼͷ࠷໬ਪఆ ࣍ʹෛͷର਺໬౓ (3.11) Λ࠷খʹ͢Δ w ΛٻΊΔͨΊʹ (3.11) ͷ w

    ʹର͢Δޯ഑ΛٻΊΔͱҎԼͷΑ͏ʹͳΔɻ(PRML ͷԋश 4.13 ࢀর) ∇E(w) = N ∑ n=1 (yn − tn )ϕ(xn ) (3.12) ͜ͷޯ഑ͷܗ͸ਖ਼ղϥϕϧ tn ͱ༧ଌ஋ yn ͷࠩ (ͭ·Γޡࠩ) ͱجఈؔ ਺ϕΫτϧ ϕ(xn ) ͷ࿨ͷܗΛ͍ͯ͠Δɻ 28 / 47
  27. 3-2. ϩδεςΟοΫճؼͷ࠷໬ਪఆ ͜ͷޯ഑ ∇E(w) Λθϩʹ͢Δ w ΛղੳతʹٻΊΔ͜ͱ͸Ͱ͖ͳ͍ɻ ͦͷཧ༝͸༧ଌ஋ y =

    σ(wTϕ(x)) ͕ϩδεςΟοΫؔ਺Λ׆ੑԽؔ਺ ʹ͔࣋ͭΒͰ͋Δɻ ͜ͷΑ͏ʹޯ഑ ∇E(w) Λθϩʹ͢Δ w ΛղੳతʹٻΊΔ͜ͱ͕Ͱ͖ ͳ͍࣌͸ޯ഑߱Լ๏Λ༻͍Δ͜ͱ͕͋Δɻ(χϡʔϥϧωοτͰ΋͜ͷ ํ๏͕Α͘༻͍ΒΕΔɻ) ޯ഑߱Լ๏Ͱ͸ɺ·ͣϥϯμϜʹܾΊͨύϥϝʔλͷॳظ஋Λ w(0) ͱ ͠ɺޡࠩؔ਺ͷޯ഑Λ༻͍ͯύϥϝʔλΛҎԼͷΑ͏ʹߋ৽͢Δɻ w(1) = w(0) − η∇E(w(0)) (3.13) ͜͜Ͱ η > 0 ͸ֶशύϥϝʔλͱݺͿɻ ͜ΕΛ܁Γฦ͢͜ͱͰύϥϝʔλ͕ޯ഑ ∇E(w) ͕খ͘͞ͳΔํ޲ʹߋ ৽͞ΕɺE(w) Λ࠷খʹ͢Δύϥϝʔλʹऩଋ͢Δɻ 29 / 47
  28. ϕΠζਪఆʹ͍ͭͯ ͜Ε·Ͱ (࠷໬ਪఆ) Ͱ͸ɺ໬౓ؔ਺Λ࠷େʹ͢ΔΑ͏ͳύϥϝʔλ w Λ఺ਪఆ͖ͯͨ͠ɻ ϕΠζਪఆͰ͸ɺڭࢣσʔλΛ༻͍ͯύϥϝʔλ w ͷ֬཰෼෍ (఺Ͱͳ

    ͘෯Λ΋ͭɺࣄޙ෼෍ͱݺ͹ΕΔ) ΛٻΊΔɻ ͦͷࣄޙ෼෍Λ༻͍ͯɺະ஌ͷσʔλͷೖྗ x ͕༩͑ΒΕͨ࣌ͷग़ྗ t ͷ༧ଌ෼෍ p(t|x, t, X) ΛٻΊΔɻ ͜ͷ༧ଌ෼෍ p(t|x, t, X) ͸ɺະ஌ͷσʔλʹର͢Δ৚݅෇͖֬཰ p(t|x, w) ͱࣄޙ෼෍ p(w|t, X) Λ༻͍ͯ p(t|x, t, X) = ∫ p(t|x, w)p(w|t, X) dw (3.14) ͱॻ͚Δɻ(PRML 1.68 ࣜࢀর) ϕΠζਪఆʹ͍ͭͯ͸ɺPRML 1.2.3 ࢀরɻ 30 / 47
  29. 3-3. ϕΠζϩδεςΟοΫճؼ ࣍͸ϩδεςΟοΫճؼΛϕΠζతʹѻ͏͜ͱΛߟ͑Δɻ ϕΠζਪఆͰ͸ະ஌ͷೖྗ x ʹର͢Δग़ྗ t ͷ༧ଌ෼෍ p(t|x, t,

    X) Λ ٻΊΔ͜ͱ͕໨తͱͳΔɻ (3.14) ΑΓɺͦͷ༧ଌ෼෍ p(t|x, t, X) ͸ॏΈ w ͷੵ෼ͰҎԼͷΑ͏ʹ ͔͚Δɻ p(t|x, t, X) = ∫ p(t|x, w)p(w|t, X) dw (3.15) ͜͜Ͱɺp(t|x, w) ͸໬౓ؔ਺Ͱ͋Γɺp(w|t, X) ͸ύϥϝʔλͷࣄޙ෼ ෍Ͱ͋Δɻ ಛʹࠓճ͸ೋ஋෼ྨΛߟ͍͑ͯΔͷͰɺ֬཰ p(C1 |x, t, X) = ∫ p(C1 |x, w)p(w|t, X) dw (3.16) ͚ͩΛੵ෼ͯ͠ٻΊͯɺp(C2 |x, t, X) ͸ p(C2 |x, t, X) = 1 − p(C1 |x, t, X) (3.17) ͷΑ͏ʹن֨Խ৚͔݅ΒٻΊΔ͜ͱΛߟ͑Δɻ 31 / 47
  30. 3-3. ϕΠζϩδεςΟοΫճؼ ͨͩ͠ɺੵ෼ (3.16) Λղੳతʹղ͘ͷ͸ෆՄೳͰ͋Δɻ(͜Ε͸ϩδε ςΟοΫγάϞΠυؔ਺ͷӨڹͰ͋Δ) ͕ͨͬͯ͠ɺੵ෼ΛۙࣅతʹٻΊΔ͜ͱΛߟ͑Δɻ ࠓճ͸ (PRML 4

    ষͰ͸) ϥϓϥεۙࣅΛ༻͍ͯੵ෼ΛۙࣅతʹٻΊͯ ͍Δɻ ۩ମతʹ͸ (3.16) ͷύϥϝʔλͷࣄޙ෼෍ p(w|t, X) ʹϥϓϥεۙࣅ Λద༻ͯ͠ɺΨ΢ε෼෍ʹۙࣅ͢Δɻ ͜͜Ͱϥϓϥεۙࣅͷઆ໌Λগ͠ߦ͏ɻ 32 / 47
  31. 3-3. ϕΠζϩδεςΟοΫճؼ ·ͣ͸֬཰ม਺͕Ұ࣍ݩͷม਺ z ͷ৔߹Λߟ͑ɺҎԼͷΑ͏ͳ֬཰෼෍ p(z) Λߟ͑Δɻ p(z) = 1

    Z f(z) (3.18) ͜͜ͰɺZ ͸ҎԼͰఆٛ͞ΕΔن֨Խఆ਺Ͱ͋Δɻ Z = ∫ f(z) dz (3.19) ϥϓϥεۙࣅͷ໨త͸෼෍ p(z) ΛϞʔυ (dp(z)/dz = 0 ͱͳΔ z) Λத ৺ͱ͢ΔΨ΢ε෼෍ʹۙࣅ͢Δ͜ͱͰ͋Δɻ ·ͣ͸Ϟʔυ z = z0 Λݟ͚ͭΔɻϞʔυ͸ (3.18) ΑΓ df(z) dz z=z0 = 0 (3.20) ͳΔ z0 Ͱ͋Δɻ 33 / 47
  32. 3-3. ϕΠζϩδεςΟοΫճؼ Ϟʔυ͕ٻ·ͬͨΒɺؔ਺ ln f(z) Λ z = z0 पΓͰҎԼͷΑ͏ʹςΠ

    ϥʔల։ͷ 2 ࣍·ͰͰۙࣅ͢Δɻ ln f(z) ∼ ln f(z0 ) − 1 2 A(z − z0 )2 (3.21) ͜͜Ͱɺ A = − d2 dz2 ln f(z) z=z0 (3.22) Ͱ͋Δɻ ͜͜Ͱɺ(3.20) ʹΑΓ (3.21) ͷӈลͰҰ࣍ͷ߲͕ଘࡏ͠ͳ͍ɻ (3.21) ͷ྆ลͷࢦ਺ΛͱΔͱ f(z) ∼ f(z0 ) exp { − 1 2 A(z − z0 )2 } (3.23) ͱͳΔɻ 34 / 47
  33. 3-3. ϕΠζϩδεςΟοΫճؼ ن֨ԽΛ͢Δͱɺ෼෍ p(z) ͸ p(z) ∼ ( A 2π

    )1/2 exp { − 1 2 A(z − z0 )2 } (3.24) ͱۙࣅͰ͖Δɻ͜Ε͕ϥϓϥεۙࣅͰ͋Δɻ ͨͩ͠஫ҙ఺ͱͯ͠ɺA > 0 Ͱͳ͍ͱΨ΢ε෼෍͕ఆٛͰ͖ͳ͍ɻ 35 / 47
  34. 3-3. ϕΠζϩδεςΟοΫճؼ ࣍͸Ұ࣍ݩͷ֬཰ม਺͔ΒɺϕΫτϧʹ֦ு͠Α͏ɻ ͭ·ΓɺҎԼͷ֬཰෼෍ p(z) Λఆٛ͢Δɻ p(z) = 1 Z

    f(z) (3.25) ͜͜Ͱɺ Z = ∫ f(z) dz (3.26) Ͱ͋Δɻ Ұ࣍ݩͷ֬཰ม਺ͱಉ͡Α͏ʹޯ഑ ∇f(z) ͕θϩʹͳΔ఺ z0 Λٻ ΊΔɻ 36 / 47
  35. 3-3. ϕΠζϩδεςΟοΫճؼ Ϟʔυ͕ٻ·ͬͨΒɺln f(z) Λ z0 पΓͰςΠϥʔల։Ͱۙࣅ͢Δɻ ln f(z) ∼

    ln f(z0 ) − 1 2 (z − z0 )TA(z − z0 ) (3.27) ͜͜ͰɺA ͸ҎԼͰఆٛ͞ΕΔ M × M ͷϔοηߦྻͰ͋Δɻ A = −∇∇ ln f(z) z=z0 (3.28) ࣍ʹ (3.27) ͷ྆ลͷࢦ਺ΛͱΔͱҎԼͷΑ͏ʹͳΔɻ f(z) ∼ f(z0 ) exp { − 1 2 (z − z0 )TA(z − z0 ) } (3.29) ͜ΕΑΓن֨ԽΛ͢Δͱɺ෼෍ p(z) ͸ p(z) ∼ |A|1/2 (2π)M/2 exp { − 1 2 (z − z0 )TA(z − z0 ) } = N(z|z0 , A−1) (3.30) ͱΨ΢ε෼෍ʹۙࣅͰ͖Δɻ 37 / 47
  36. 3-3. ϕΠζϩδεςΟοΫճؼ Ҏ্Ͱઆ໌ͨ͠ϥϓϥεۙࣅΛ༻͍ͯҎԼͷੵ෼ (3.16) Λۙࣅ͍ͨ͠ɻ p(C1 |x, t, X) =

    ∫ p(C1 |x, w)p(w|t, X) dw (3.31) ·ͣɺࣄޙ෼෍ p(w|t, X) ΛٻΊΔͨΊʹࣄલ෼෍Λಋೖ͢Δɻ p(w) = N(w|m0 , S0 ) (3.32) (3.18) ΑΓɺ໬౓ؔ਺͸ p(t|X, w) ͸ p(t|X, w) = N ∏ n=1 ytn n (1 − yn )1−tn (3.33) Ͱ͋ͬͨɻ 38 / 47
  37. 3-3. ϕΠζϩδεςΟοΫճؼ ͜ΕΑΓɺࣄޙ෼෍ p(w|t, X) ͸ϕΠζͷఆཧΑΓɺҎԼͰ͋Δɻ p(w|t, X) ∝ p(w)p(t|X,

    w) (3.34) ͱͳΔͷͰɺln p(w|t, X) ͸ҎԼͱͳΔɻ ln p(w|t, X) = − 1 2 (w − m0 )TS−1 0 (w − m0 ) + N ∑ n=1 { tn ln yn + (1 − tn ) ln (1 − yn ) } + const. (3.35) ͜ͷࣄޙ෼෍ͷର਺ ln p(w|t, X) Λ࠷େʹ͢Δύϥϝʔλ wMAP Λ (ͨ ͱ͑͹ޯ഑߱Լ๏ͳͲͰ) ٻΊͯɺͦͷ఺ wMAP ͰͷϔοηߦྻΛٻΊ ΔͱɺҎԼͷΑ͏ʹͳΔɻ S−1 N = − ∇∇ ln p(w|t, X) w=wMAP =S−1 0 + N ∑ n=1 yn (1 − yn )ϕn ϕT n w=wMAP (3.36) 39 / 47
  38. 3-3. ϕΠζϩδεςΟοΫճؼ ΑͬͯɺϥϓϥεۙࣅΛ༻͍Δͱࣄޙ෼෍ p(w|t, X) ͸ҎԼͷΑ͏ʹۙ ࣅͰ͖Δɻ p(w|t, X) ∼

    N(w|wMAP , SN ) (3.37) ͜ΕΑΓɺ(3.31) ͷੵ෼͸ҎԼͷΑ͏ʹۙࣅͰ͖Δɻ p(C1 |x, t, X) ∼ ∫ σ(wTϕ) N(w|wMAP , SN ) dw (3.38) ͜͜Ͱɺp(C1 |x, w) = σ(wTϕ) Λར༻ͨ͠ɻ ࣍ʹɺϩδεςΟοΫγάϞΠυؔ਺ΛҎԼͷΑ͏ʹॻ͖௚͢ɻ σ(wTϕ) = ∫ δ(a − wTϕ)σ(a) da (3.39) ͜͜Ͱɺδ(·) ͸σϡϥοΫͷσϧλؔ਺Ͱ͋Δɻ 40 / 47
  39. 3-3. ϕΠζϩδεςΟοΫճؼ ͜ΕΑΓɺ(3.38) ͸ҎԼͷΑ͏ʹॻ͖௚ͤΔɻ ∫ σ(wTϕ) N(w|wMAP , SN )

    dw = ∫ ∫ δ(a − wTϕ)σ(a) N(w|wMAP , SN ) da dw = ∫ ∫ δ(a − wTϕ) N(w|wMAP , SN ) dw σ(a) da = ∫ p(a) σ(a) da (3.40) ͜͜Ͱɺ p(a) = ∫ δ(a − wTϕ) N(w|wMAP , SN ) dw (3.41) Ͱ͋Δɻ 41 / 47
  40. 3-3. ϕΠζϩδεςΟοΫճؼ ੵ෼ (3.41) ʹ͓͍ͯɺϕ ʹฏߦͳ͢΂ͯͷํ޲ͷ w ੵ෼͸ͦΕΒͷύ ϥϝʔλʹઢܗ੍໿Λ༩͑ɺ·ͨ ϕ

    ʹ௚ߦ͢Δ͢΂ͯͷํ޲ͷ w ੵ෼ ͸Ψ΢ε෼෍ N(w|wMAP , SN ) ͷपลԽΛ༩͑Δɻ ͨͱ͑͹ɺw = (w1 , w2 )T ͱ͠ɺϕ = (ϕ, 0)T Ͱ͋Δͱ͖Λߟ͑Δͱɺ ੵ෼ (3.41) ͸ҎԼͷΑ͏ʹ͔͚Δɻ p(a) = ∫ ∫ δ(a − w1 ϕ) N(w|wMAP , SN ) dw1 dw2 = ∫ δ(a − w1 ϕ) [ ∫ N(w|wMAP , SN ) dw2 ] dw1 (3.42) (??) ΑΓɺΨ΢ε෼෍ΛपลԽͨ͠पล෼෍͸࠶ͼΨ΢ε෼෍Ͱ͋Δ ͜ͱ͕Θ͔͍ͬͯΔͷͰɺϕ ʹ௚ߦ͢Δ w2 ํ޲ͷੵ෼͸Ψ΢ε෼෍ͷ पลԽΛ༩͑ɺͦͷपลԽ͞ΕͨΨ΢ε෼෍͸ N(w1 |(wMAP )1 , (SN )11 ) ͱͳΔɻ 42 / 47
  41. 3-3. ϕΠζϩδεςΟοΫճؼ ·ͨɺw1 ͷੵ෼Λ͢Δͱɺੵ෼ (3.41) ͸ҎԼͷΑ͏ʹͳΔɻ p(a) = ∫ δ(a

    − w1 ϕ) N(w1 |(wMAP )1 , (SN )11 ) dw1 = 1 |ϕ| N(a/ϕ|(wMAP )1 , (SN )11 ) =N(a|(ϕwMAP )1 , (ϕ2SN )11 ) (3.43) ͭ·Γɺϕ ʹฏߦͳ w1 ͷํ޲ͷੵ෼͸ w1 ʹ w1 = a/ϕ ͳΔઢܗ੍໿ Λ༩͑Δ͜ͱ͕Θ͔Δɻ ͜ΕΑΓɺp(a) ͸֬཰ม਺͕ a ͷΨ΢ε෼෍ʹͳΔ͜ͱ͕Θ͔Δɻ 43 / 47
  42. 3-3. ϕΠζϩδεςΟοΫճؼ Ψ΢ε෼෍͸ฏۉͱ෼ࢄ͕ܾ·Ε͹ɺܗ͕Ұҙʹఆ·Γɺฏۉ µa ͱ෼ ࢄ σ2 a ͸ҎԼͷΑ͏ʹͳΔɻ µa

    = ∫ p(a)a da = ∫ ∫ aδ(a − wTϕ) N(w|wMAP , SN ) dwda = ∫ wTϕ N(w|wMAP , SN ) dw = wT MAP ϕ (3.44) σ2 a = ∫ p(a)(a2 − µ2 a ) da = ∫ ∫ (a2 − µ2 a )δ(a − wTϕ) N(w|wMAP , SN ) dwda = ∫ ((wTϕ)2 − (wT MAP ϕ)2) N(w|wMAP , SN ) dw =ϕT [ ∫ (wwT − wMAP wT MAP ) N(w|wMAP , SN ) dw ] ϕ =ϕTSN ϕ (3.45) 44 / 47
  43. 3-3. ϕΠζϩδεςΟοΫճؼ ͢Δͱɺ༧ଌ෼෍ p(C1 |x, t, X) ͸ (3.40) ΑΓɺҎԼͷΑ͏ʹͳΔ͜ͱ

    ͕Θ͔Δɻ p(C1 |x, t, X) ∼ ∫ σ(a)N(a|µa , σ2 a ) da (3.46) ͜͜Ͱɺµa ͱ σ2 a ͸ (3.44) ͱ (3.45) Ͱܭࢉͨ͠ฏۉͱ෼ࢄͷύϥϝʔ λͰ͋Δɻ ͜ͷੵ෼ (3.46) ΋·ͨղੳతʹੵ෼Ͱ͖ͳ͍ɻ ͦ͜ͰҎԼͷϓϩϏοτؔ਺ͷٯؔ਺ Φ(a) Λಋೖ͢Δɻ Φ(a) = 1 2 { 1 + erf ( a √ 2 )} (3.47) ͜͜Ͱɺޡࠩؔ਺ erf(a) ͸ҎԼͰఆٛ͞ΕΔɻ erf(a) = 2 √ π ∫ a 0 exp (−θ2) dθ (3.48) 45 / 47
  44. 3-3. ϕΠζϩδεςΟοΫճؼ ϓϩϏοτؔ਺ͷٯؔ਺ Φ (√ π 8 a ) ʹΑͬͯϩδεςΟοΫγάϞΠ

    υؔ਺ σ(a) Λۙࣅ͢Δ͜ͱ͕Ͱ͖Δɻ ҎԼ͸ϩδεςΟοΫγάϞΠυؔ਺ σ(a)(੺ͷ࣮ઢ) ͱϓϩϏοτؔ ਺ͷٯؔ਺ Φ (√ π 8 a ) (੨ͷ఺ઢ) Λൺֱͨ͠ਤͰ͋Δɻ 46 / 47
  45. 3-3. ϕΠζϩδεςΟοΫճؼ ͞ΒʹϓϩϏοτؔ਺ͷٯؔ਺ʹ͸ҎԼͷੑ࣭͕͋Δɻ(PRML ͷԋश 4.26 ࢀর) ∫ Φ(λa)N(a|µ, σ2) da

    = Φ ( µ (λ−2 + σ2)1/2 ) (3.49) ͜ΕΒͷੑ࣭Λ༻͍ͯɺੵ෼ (3.46) ΛҎԼͷΑ͏ʹۙࣅͯ͠ٻΊΔɻ p(C1 |x, t, X) ∼ ∫ σ(a)N(a|µa , σ2 a ) da ∼ ∫ Φ (√ π 8 a ) N(a|µa , σ2 a ) da =Φ ( µa (8/π + σ2 a )1/2 ) ∼ σ (√ 8 π µa (8/π + σ2 a )1/2 ) =σ ( µa (1 + πσ2 a /8)1/2 ) (3.50) ͜͜Ͱɺµa ͱ σ2 a ͸ (3.44) ͱ (3.45) Ͱ͋Δɻ 47 / 47