NxでMNISTの手書き数字画像分類を試す / Training MNIST Datasets with Nx

Transcript

1. ܀ྛ
   ݈ଠ࿠ʢ(.0ϖύϘגࣜձࣾɺ๺཮ઌ୺Պֶٕज़େֶӃେֶʣ /FSWFT+1
   
   /YΛ৮ͬͯΈΔճʢ೥݄೔ʣ NxͰMNISTͷखॻ͖਺ࣈը૾෼ྨΛࢼ͢

2. ܀ྛ
   ݈ଠ࿠
   BLB
   ͋ΜͪΆ
   IUUQTLFOUBSPLVSJCBZBTIJDPN
   ɾ(.0ϖύϘגࣜձࣾऔక໾$50
   ɾҰൠࣾஂ๏ਓ೔ຊ$50ڠձཧࣄ
   ๺཮ઌ୺Պֶٕज़େֶӃେֶʢ+"*45ʣത ࢜લظ՝ఔࡏֶதͷࣾձਓֶੜͰ΋͋ Δɻ
   *P5ؔ࿈ͷݚڀΛ४උ͍ͯ͠Δͱ͜Ζʢ ݄ʹ/FSWFT͕ग़ͯ͘Δݚڀใࠂ࿦จʹͭ ͍ͯൃද͠·͢ʣɻ

   ࣗݾ঺հ 2

3. /Y
   /VNFSJDBM
   &MJYJS
   JT
   OPX
   QVCMJDMZ
   BWBJMBCMF
   
   %BTICJU
   #MPH
   IUUQTEBTICJUDPCMPHOYOVNFSJDBMFMJYJSJTOPXQVCMJDMZBWBJMBCMF

4. *OUSPEVDJOH
   /Y
   
   +PTÉ
   7BMJN
   c
   -BNCEB
   %BZT
   
   IUUQTZPVUVCFG1,.N+Q"(8D

5. ಈըΛ؍ͯΔ͚ͩͰ͸Θ͔Βͳ͍ͷͰ ϥΠϒίʔσΟϯάΛࣸܦͰ࠶ݱͨ͠

6. +OOOY
   
   +PTÉ`T
   /FVSBM
   /FUXPSL
   XJUI
   /Y
   IUUQTHJUIVCDPNLFOUBSPKOOOY

7. ͜Μͳײ͡ͰKeras෩ʹࢼͤ·͢ 
   ./*45σʔλͷಡΈࠐΈ
   [x_train, y_train, x_test, y_test] = Jnnnx.MNIST.Dataset.load_data() 
   σʔλͷ੔ܗͱਖ਼نԽ
   x_train

   = x_train |> Nx.reshape({60000, 28*28}, names: [:batch, :input]) |> Nx.divide(255) x_test = x_test |> Nx.reshape({10000, 28*28}, names: [:batch, :input]) |> Nx.divide(255) 
   POFIPUFODPEJOH
   y_train = y_train |> Jnnnx.Utils.to_categorical(10, names: [:batch, :output]) y_test = y_test |> Jnnnx.Utils.to_categorical(10, names: [:batch, :output]) 
   τϨʔχϯάσʔλΛ༻ֶ͍ͯश
   params = Jnnnx.fit(x_train, y_train, epoch: 5, batch_size: 50, learning_rate: 0.01) 
   ςετσʔλΛ༻͍ͯධՁ
   score = Jnnnx.evaluate(params, x_test, y_test) IO.puts("Accuracy: #{Nx.to_scalar(score)}")

8. ૉͷ&MJYJS
   
   $16ʢ&9-"Λ༻͍ͳ͍ʣͰ࣮ߦͨ݁͠Ռ˞
   ֶश݁ՌʢΤϙοΫ
   5ɺֶश཰
   0.01ɺֶशʹཁͨ࣌ؒ͠: ൒೔͙Β͍ʣ ˞
   &9-"
   
   $16΋ಈ͔ͯ͠Έ͕ͨɺܻ୆Ͱ଎͘ͳΔʢ࣍ϖʔδʣͱ͸͍͑ݩ͕஗ա͗ΔͷͰಉֶ͡शΛ΋͏Ұ౓΍Δ͜ͱ͸͠ͳ͔ͬͨɻ

9. 4PGUNBYؔ਺ͷ࣮ߦํࣜ͝ͱͷϕϯνϚʔΫ݁Ռ
   IUUQTHJUIVCDPNFMJYJSOYOYUSFFNBJOOYOVNFSJDBMEFGJOJUJPOT

10. ˔ +PTÉͷϥΠϒίʔσΟϯάಈըΛ؍ͳ͕Βࣸܦͨ͠
    ˔ ػցֶशϥΠϒϥϦͷΑ͏ʹ࢖͑ΔΑ͏ʹ੔ཧͨ͠
    ˓ ࣸܦͨ͠ίʔυΛϥΠϒϥϦͬΆ͍ϑΝΠϧߏ଄Ͱ഑ஔ
    ˓ ϋΠύʔύϥϝλΛؔ਺ͷҾ਺ͱͯ͠౉ͤΔΑ͏ʹͨ͠
    ˔ ./*45ͷσʔληοτΛऔಘ͢ΔϞδϡʔϧΛ௥Ճͨ͠
    

    ˔ ֶशͨ͠ϞσϧΛɺςετσʔλʹΑͬͯධՁ͢Δؔ਺Λ௥Ճͨ͠
    ˠ
    ಈըͰσϞͯͨ͠ίʔυͷݩʹͳ͍ͬͯΔ΋ͷͱࢥΘΕΔ΋ͷ͕ FYMBͷ΄͏ͷFYBNQMFTʹ͋ͬͨʂ˞ ΍ͬͨ͜ͱ ˞
    IUUQTHJUIVCDPNFMJYJSOYOYCMPCNBJOFYMBFYBNQMFTNOJTUFYT

11. ˔ ݱঢ়Ͱ͸ϨΠϠʔͷߏ੒ɺ׆ੑԽؔ਺ɺଛࣦؔ਺౳ΛܾΊଧͪʹ͠ ͍ͯΔ͕ɺࣗ༝ʹ૊Έ߹ΘͤΒΕΔΑ͏ʹ͢Δ͜ͱ
    ˓ ͦͷ͋ͨΓ·Ͱ΍Δͱ΋͏গ͠ϥΠϒϥϦͬΆ͘ͳΔ
    ˓ ͍·͸୯ʹॲཧΛͦΕͬΆ͘ݟ͑ΔΑ͏ʹ·ͱΊ͚ͨͩ
    ˔ &9-"Λ࢖ͬͯ(16Ͱܭࢉ͢Δ͜ͱ˞
    ·ͩ΍ͬͯͳ͍͜ͱ

    ˞
    +FUTPO
    /BOP
    (#Ͱࢼ͔͕ͨͬͨ͠ɺϕλϕλ৮ͬͨΓ͍͔ͨͤ͠ىಈ͠ͳ͘ͳͬ

12. ͜Ε͔Βඞཁͳ΋ͷ͕੝Γͩ͘͞Μʂ
    IUUQTZPVUVCFG1,.N+Q"(8D U

13. ˔ ݱঢ়͸ɺςϯιϧͷܭࢉ΍ࣗಈඍ෼౳ͷɺσΟʔϓϥʔχϯάΛ͢Δ ্ͰجຊͱͳΔϏϧσΟϯάϒϩοΫ͕Ͱ͖ͨͱ͜Ζ
    ˔ 5FOTPSGMPX,FSBTɺ1Z5PSDIɺTDJLJUMFBSOͷΑ͏ͳػցֶशϑϨʔ ϜϫʔΫ͕͋Δͱ͍͍ͳ͋
    ˠ
    ͦΜͳؾ࣋ͪ΋͋ͬͯࠓճɺࡶʹࢼͯ͠ΈͨΓͨ͠ͷͰͨ͠
    ˠ
    ʰθϩ͔Β࡞Δ%FFQ
    -FBSOJOH
    ⁠ʕϑϨʔϜϫʔΫฤʱΛಡΈͳ ͕Βࢼ͠ʹ࡞Γ࢝ΊͯΈ͚ͨͲ͏·͍͜ͱઃܭͰ͖ͳ͍ͯͬͨ͘Μ͋ ͖ΒΊ·ͨ͠ʢؔ਺ܕݴޠ೴ʹͳΓ͖Εͯͳ͍ʜʜʣ

    ࠓޙ΁ͷظ଴