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

人工知能と機械学習 / AI and ML

Yukino Baba
September 26, 2019

人工知能と機械学習 / AI and ML

Yukino Baba

September 26, 2019
Tweet

More Decks by Yukino Baba

Other Decks in Education

Transcript

  1. /39 ೔ৗͷதͷਓ޻஌ೳ w εϚʔτεϐʔΧʔɿ
 ਓؒͷԻ੠ࢦࣔʹै͍༷ʑͳಈ࡞Λߦ͏ w ϩϘοτ૟আػɿ
 ো֐෺Λආ͚ͳ͕Β෦԰શମΛ૟আ͢Δ w إೝূɿ


    Χϝϥલͷਓ෺͕ొ࿥͞ΕͨਓͱಉҰਓ෺͔Ͳ͏͔Λ൑ఆ w ίϯςϯπɾ঎඼ਪનɿ
 Ӿཡཤྺ΍ߪങཤྺʹ΋ͱ͖ͮϢʔβ͕޷ΉΞΠςϜΛఏࣔ զʑͷ೔ৗͷ͞·͟·ͳ৔໘Ͱਓ޻஌ೳ͕׆༻͞Ε͍ͯΔ 8
  2. /39  ݱࡏͷਓ޻஌ೳʹͰ͖Δ͜ͱɿಛԽܕWT൚༻ܕ w ਓ޻஌ೳ͸ʮಛԽܕʯͱʮ൚༻ܕʯͷೋछྨʹେผ͞ΕΔ w ಛԽܕਓ޻஌ೳɿ
 ͋Δಛఆͷঢ়گ΍໰୊ʹ͓͍ͯ஌తʹ;Δ·͏ਓ޻஌ೳ w ൚༻ܕਓ޻஌ೳɿ


    ਓؒͱಉ༷ʹ༷ʑͳঢ়گɾ໰୊ʹ͓͍ͯ஌తͳ;Δ·͍͕Ͱ͖Δਓ޻஌ೳ w ݱࡏͷਓ޻஌ೳ͸ಛԽܕਓ޻஌ೳͰ͋Γ
 ൚༻ܕਓ޻஌ೳ͸ະ࣮ͩݱ͞Ε͍ͯͳ͍
 ݱࡏͷਓ޻஌ೳ͸ಛఆͷ͜ͱ͚ͩͰ͖ΔಛԽܕਓ޻஌ೳ 10
  3. /39 ݱࡏͷਓ޻஌ೳʹͰ͖Δ͜ͱɿը૾ॲཧ w ը૾ೝࣝɿ
 ࣸਅʹ͍ࣸͬͯΔ΋ͷΛೝࣝ͢Δ w ෺ମݕग़ɿ
 ࣸਅͷͲ͜ʹԿ͕͍ࣸͬͯΔͷ͔Λೝࣝ͢Δ w ը૾ੜ੒ɿ


    ຊ෺ͷࣸਅͷΑ͏ͳը૾Λਓ޻తʹ࡞Γग़͢ ಛԽܕਓ޻஌ೳͰ΋༷ʑͳ͜ͱ͕࣮ݱͰ͖Δ 11 ग़యɿIUUQDTOTUBOGPSEFEVTMJEFTDTO@@MFDUVSFQEG
 ग़యɿIUUQTBSYJWPSHBCT ຊ෺ͷࣸਅ ਓ޻తͳࣸਅ (*) (*)
  4. /39 ݱࡏͷਓ޻஌ೳʹͰ͖Δ͜ͱɿݴޠॲཧ w จॻ෼ྨɿ
 จॻΛಛఆͷΧςΰϦʹ෼ྨ͢Δ w ػց຋༁ɿ
 ͋ΔݴޠͰॻ͔ΕͨจষΛผͷݴޠʹ຋༁͢Δ w ৘ใݕࡧɿ


    ಛఆͷ৘ใ͕ܝࡌ͞Ε͍ͯΔ΢ΣϒϖʔδΛݟ͚ͭΔ w ࣭໰Ԡ౴ɿ
 จষͰ༩͑ΒΕ࣭ͨ໰ʹର͢Δ౴͑Λฦ͢ ಛԽܕਓ޻஌ೳͰ΋༷ʑͳ͜ͱ͕࣮ݱͰ͖Δ 12
  5. /39 ػցֶशͱ͸ ೖྗͱग़ྗΛରԠ͚ͮΔϧʔϧΛֶश͢Δ 20 The quick brown fox jumps over

    the lazy dog ૉૣ͍஡৭ͷޅ͸ ͷΖ·ͳݘΛඈͼ ӽ͑Δ “Hello, world!” ೖྗ ग़ྗ ݈߁ϦεΫɿߴ ग़యɿIUUQDTOTUBOGPSEFEVTMJEFTDTO@@MFDUVSFQEG
 ग़యɿIUUQTNBHFOUBUFOTPSqPXPSHOTZOUIGBTUHFO จষ จষ ը૾ ෺ମͱҐஔ Ի੠ จষ ױऀ৘ใ ݈߁ϦεΫ (*) (**)
  6. /39 ػցֶशͷ͘͠Έɿ܇࿅σʔλ w ྫ୊ɿʮजᙾ͕ྑੑ͔ѱੑ͔ʯ൑ఆ͢ΔͨΊͷϧʔϧΛֶशͤ͞Δ w ڭࡐͱͯ͠༻͍ΔσʔλΛ܇࿅σʔλͱݺͿ w ֶशͨ͠ϧʔϧΛ༻͍ͯ
 ܇࿅σʔλʹͳ͍जᙾ͕ྑੑ͔ѱੑ͔Λਖ਼͘͠൑ఆͰ͖ΔΑ͏ʹ͍ͨ͠
 ೖग़ྗྫΛ܇࿅σʔλͱͯ͠༻͍ͯϧʔϧΛֶश͢Δ

    22 जᙾը૾ͷग़యɿW. N. Street et al. , Nuclear feature extraction for breast tumor diagnosis. Proc. of the International Symposium on Electronic Imaging, 1993. ྑੑ ѱੑ ྑੑ ѱੑ ܇࿅σʔλ ೖྗʢजᙾʣΛ
 ग़ྗʢྑੑPSѱੑʣʹ
 ରԠ͚ͮΔ
 ϧʔϧ ֶश
  7. /39 ػցֶशͷ͘͠Έɿ܇࿅σʔλ ܇࿅σʔλ͸਺஋Ͱදݱ͞Εͳ͍ͱ͍͚ͳ͍ 24 211 220 223 … 213 217

    217 220 … 210 214 217 216 … 205 … … … … … 216 217 214 … 181 The quick brown fox jumps over the lazy dog aardvark … brown … dog … zzzat 0 … 1 … 1 … 0
  8. /39 ػցֶशͷ͘͠Έɿ܇࿅σʔλ ܇࿅σʔλ͸਺஋Ͱදݱ͞Εͳ͍ͱ͍͚ͳ͍ 25 ܇࿅σʔλͷྫ ೖྗ ग़ྗ େ͖͞ Ͱ͜΅͜౓ जᙾ

    0.252 0.014 ྑੑ जᙾ 0.327 0.340 ྑੑ जᙾ 1.000 0.371 ѱੑ जᙾ 0.223 0.369 ѱੑ ʜ ʜ ʜ ʜ ਤࣔ େ͖͞ Ͱ͜΅͜౓ ྑੑͷजᙾ ѱੑͷजᙾ जᙾ जᙾ
  9. /39 ػցֶशͷ͘͠Έɿϧʔϧ ϧʔϧ΋਺ࣜͰදݱ͞Εͳ͍ͱ͍͚ͳ͍ 27 େ͖͞ Ͱ͜΅͜౓ ྑੑͷजᙾ ѱੑͷजᙾ [ʾͷྖҬ [ͷྖҬ

     ͷ৔߹ɿ
  େ͖͞ Ͱ͜΅͜౓ w1 = 1, w2 = − 1, b = 0 z = −  େ͖͞ Ͱ͜΅͜౓ z = w1 × + w2 × + b;
  10. /39 ػցֶशͷ͘͠Έɿϧʔϧ X X Cͷ஋͕มΘΔͱϧʔϧ΋มΘΔ 28 େ͖͞ Ͱ͜΅͜౓ ྑੑͷजᙾ ѱੑͷजᙾ

    [ʾͷྖҬ [ͷྖҬ  ͷ৔߹ɿ
  େ͖͞ Ͱ͜΅͜౓ w1 = 1, w2 = 1, b = − 1 z = + −1  େ͖͞ Ͱ͜΅͜౓ z = w1 × + w2 × + b;
  11. /39 ػցֶशͷ͘͠Έɿϧʔϧ ܇࿅σʔλʹ౰ͯ͸·Δྑ͍ϧʔϧΛֶश͍ͨ͠ 29 େ͖͞ Ͱ͜΅͜౓ େ͖͞ Ͱ͜΅͜౓  w1

    = 1, w2 = − 1, b = 0  w1 = 1, w2 = 1, b = − 1 ϧʔϧ͕౰ͯ͸·Βͳ͍जᙾ ӈͷϧʔϧͷํ͕ɺ܇࿅σʔλʹ౰ͯ͸·Δ
  12. /39 ػցֶशͷ͘͠Έɿଛࣦ ϧʔϧͷྑ͠ѱ͠Λ఺਺Ͱද͢ 30 ଛࣦɿେ ଛࣦɿখ େ͖͞ Ͱ͜΅͜౓ େ͖͞ Ͱ͜΅͜౓

    w ଛࣦɿϧʔϧͷѱ͞Λ఺਺Ͱදͨ͠΋ͷ w ܇࿅σʔλʹ౰ͯ͸·Βͳ͍ϧʔϧ͸ଛࣦ͕େ͖͘ͳΔ
  13. /39 ػցֶशͷ͘͠Έɿଛࣦ w ʮޡ൑ఆͷ݅਺ʯΛଛࣦͱͯ͠࢖͏ͱϧʔϧͷࡉ͔͍ҧ͍͕Θ͔Βͳ͍ w ϧʔϧʹʮѱੑͷ֬཰ʯΛग़ྗͤ͞ɺͦΕʹ΋ͱ͖ͮଛࣦΛࢉग़ ϧʔϧͷྑ͠ѱ͠Λ఺਺Ͱද͢ 31 ૯࿨Λϧʔϧͷଛࣦͱ͢Δ ଛࣦɿ

    ܇࿅σʔλ [ ѱੑͷ֬཰
 ଛࣦ ೖྗ ग़ྗ େ͖͞ Ͱ͜΅͜౓ जᙾ 0.252 0.014 ྑੑ -0.542 0.336 0.410 जᙾ 0.327 0.340 ྑੑ -0.172 0.457 0.611 जᙾ 1.000 0.371 ѱੑ 1.208 0.770 0.261 जᙾ 0.223 0.369 ѱੑ -0.348 0.414 0.882 ʜ ʜ ʜ ʜ a = σ(z)
  14. /39 ػցֶशͷ͘͠Έɿ൚Խೳྗ w ֶशͷͦ΋ͦ΋ͷ໨త͸
 ʮ܇࿅σʔλʹͳ͍ະ஌ͷೖྗʹରͯ͠ਖ਼͍͠ग़ྗΛฦ͢ʯϧʔϧͷֶश w ܇࿅σʔλ΁ͷ౰ͯ͸·Γ͚ͩΛߟྀ͢Δͱ
 ϧʔϧ͕܇࿅σʔλʹ౰ͯ͸·Γ͗ͯ͢ະ஌ͷೖྗͰ͸ؒҧ͑ΔڪΕ w X

    Xͷ஋͕ۃ୺ʹେ͖͍ϧʔϧ͸ա৒ద߹͍ͯ͠ΔڪΕ͕͋ΔͷͰ
 ͦͷΑ͏ͳϧʔϧʹϖφϧςΟΛ༩͑ΔΑ͏ʹଛࣦΛઃܭ͢Δ w ʮϧʔϧ͕ະ஌ͷೖྗʹର͠ਖ਼͍͠ग़ྗΛฦ͢ೳྗʯΛ
 ൚Խೳྗͱ͍͏ ܇࿅σʔλʹ౰ͯ͸·Γ͗͢Δϧʔϧ͸൚Խೳྗ͕௿͍ 33
  15. /39 w ௚ઢͰදݱ͞ΕΔϧʔϧ͸ೖྗͱग़ྗͷରԠ͚͕ͮෳࡶͳ৔߹ʹෆे෼ w ਂ૚ֶशΛ༻͍ΔͱෳࡶͳϧʔϧΛֶशͰ͖Δ େ͖͞ Ͱ͜΅͜౓ େ͖͞ Ͱ͜΅͜౓ ୯७ͳϧʔϧ

    ػցֶशͷ͘͠Έɿਂ૚ֶश ෳࡶͳϧʔϧͷֶशʹ༻͍ΒΕΔͷ͕ਂ૚ֶश 34 େ͖͞ Ͱ͜΅͜౓ ਂ૚ֶशʹΑΔϧʔϧ
  16. /39 ػցֶशͷ࣮૷ w ਂ૚ֶशͷ1ZUIPOϥΠϒϥϦ΋༷ʑͳ΋ͷ͕ఏڙ͞Ε͍ͯΔɿ
 5FOTPS'MPX ,FSBT 1Z5PSDI $IBJOFS w (PPHMF$PMBCPSBUPSZΛ࢖͏ͱԾ૝ϚγϯΛ࢖ͬͯ


    ਂ૚ֶशϥΠϒϥϦΛ΢Σϒϒϥ΢β͔Β؆୯ʹ࢖͏͜ͱ͕Ͱ͖Δ ༷ʑͳϥΠϒϥϦ͕ެ։͞Ε͓ͯΓ؆୯ʹ࣮૷͕Ͱ͖Δ 37 https://colab.research.google.com