AWSの機械学習基盤を使ってみよう

合同勉強会 in 大都会岡山 -2017 Winter- での登壇資料
https://gbdaitokai.connpass.com/event/58025/

Transcript

1. AWSͷػցֶशج൫Λ࢖ͬͯΈΑ͏ 
   
   
   
   
   ౔
   ߹ಉษڧձ in େ౎ձԬࢁ -2017 Winter- ాத޹໌
2. "CPVU
   NF
3. wΫϥεϝιουגࣜձࣾ
   wϞόΠϧΞϓϦαʔϏε෦
   wJ04ΞϓϦΤϯδχΞ
   wαʔόʔαΠυΞϓϦΤϯδχΞ
   wαʔόʔϨε։ൃ෦
   wΞϓϦέʔγϣϯΤϯδχΞ ాத޹໌ @kongmingtrap
4. ాத޹໌ @kongmingtrap Ԭࢁग़਎
5. ాத޹໌ @kongmingtrap ෱Ԭࡏॅ
6. IUUQTDMBTTNFUIPEKQOFXTOFXP⒏DFGVLVPLB
7. IUUQTDMBTTNFUIPEKQOFXTOFXP⒏DFGVLVPLB ෱ԬҠॅʂʂʂ
8. ؓ࿩ٳ୊
9. ࠓ೥ͷػցֶश
   ϋΠϥΠτ
10. w5FOTPS'MPX-JUF
    wߴ଎͔ͭܰྔͳΞϓϦ޲͚ػցֶशϑϨʔϜϫʔΫ
    w5FOTPS'MPX
    3FTFBSDI
    $PVME
    wτϨʔχϯά͓Αͼਪ࿦ͷ྆ํΛߴ଎Խ͢Δ·ͬͨ ͘৽͍͠
    (PPHMF
    ͷΫϥ΢υ
    516 (PPHMF
    *0
    
11. w$PSF
    .-
    wֶशϞσϧ౳ΛJ04
    
    NBD04্Ͱར༻͢Δࡍʹɺ ։ൃऀ͕ઐ໳తͳ஌ࣝΛඞཁͱͤͣʹѻ͑ΔΑ͏ʹ ิॿ͢ΔϑϨʔϜϫʔΫ
    wDPSFNMUPPMT
    wػցֶशϑϨʔϜϫʔΫͰ࡞੒ֶͨ͠शϞσϧΛ $PSF
    .-Ͱར༻Ͱ͖ΔΑ͏ʹม׵ 88%$
    
12. w"84
    %FFQ-FOT SF*OWFOU
    
13. w"84
    %FFQ-FOT
    wσΟʔϓϥʔχϯάϞσϧΛػث্Ͱ௚઀࣮ߦͰ͖ ΔɺϓϩάϥϛϯάՄೳͳ৽͍͠ϏσΦΧϝϥ
    wͲΜͳεΩϧϨϕϧͷ։ൃऀͰ΋෼Ͱ%FFQ
    -FBSOJOHΛ։࢝Ͱ͖Δ SF*OWFOU
    
14. w"NB[PO
    3FLPHOJUJPO
    7JEFP
    wը૾෼ੳαʔϏε"NB[PO
    3FLPHOJUJPO͕ɺಈը Λαϙʔτ
    w෺ମɺγʔϯɺςΩετɺإͷݕग़ɺ༗໊ਓͷೝࣝ
    wܞଳి࿩ɺΧϝϥɺ*P5ϏσΦηϯαʔɺ͓ΑͼϦΞ ϧλΠϜϥΠϒετϦʔϜϏσΦॲཧ͔ΒΩϟϓ νϟʔ͞ΕͨಈըΛɺεέʔϥϒϧͰߴਫ਼౓Ͱಈը ෼ੳ͢ΔιϦϡʔγϣϯʹར༻ SF*OWFOU
    
15. "NB[PO
    4BHFNBLFS
16. wػցֶशϞσϧͷߏஙͱτϨʔχϯάͷ४ උ͕ΑΓ؆୯ʹ
    wΞϓϦέʔγϣϯʹ࠷దͳΞϧΰϦζϜͱϑϨʔϜ ϫʔΫΛબ୒
    w࠷దԽ͢ΔͨΊʹඞཁͳπʔϧ͕ἧ͍ͬͯΔ "NB[PO
    4BHF.BLFS
17. wτϨʔχϯάσʔλΛ؆୯ʹ෼ੳ͠ՄࢹԽ
    wϗετܕͷ
    +VQZUFS
    /PUFCPPLΛඋ͍͑ͯΔ
    w4
    ͷσʔλʹ௚઀઀ଓͰ͖Δ
    w"NB[PO
    %ZOBNP%#ɺ"NB[PO
    3FETIJGU͔Β ͷσʔλΛ
    4
    ʹҠಈͯͦ͠ΕΒͷσʔλΛ
    /PUFCPPLͰ෼ੳͰ͖Δ "NB[PO
    4BHF.BLFS
18. Ͳ͏มΘΔͷ͔ʁ
19. ैདྷͰ͸ʜ w(16͕ࡌ͍ͬͯΔΠϯελϯεΛ༻ҙ͢Δ
20. ैདྷͰ͸ʜ wػցֶशϑϨʔϜϫʔΫΛ౥ࡌ͍ͯ͠Δ".* Λىಈ͢Δ
21. ैདྷͰ͸ʜ wֶश༻ͷϓϩάϥϜΛ४උ͢Δ model = Sequential() model.add(Conv2D(32, (3, 3), padding='same', input_shape=X_train.shape[1:])) model.add(Activation('relu')) model.add(Conv2D(32, (3, 3))) model.add(Activation('relu')) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Dropout(0.25)) model.add(Conv2D(64, (3, 3), padding='same')) model.add(Activation('relu')) model.add(Conv2D(64, (3, 3))) model.add(Activation('relu')) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Dropout(0.25)) model.add(Flatten()) model.add(Dense(512)) model.add(Activation('relu')) model.add(Dropout(0.5)) model.add(Dense(nb_classes)) model.add(Activation('softmax')) model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
22. ैདྷͰ͸ʜ w࣮ߦ
23. ैདྷͰ͸ʜ w(16Πϯελϯεͷ༻ҙɺֶश༻ͷϓϩά ϥϜͷ࡞੒ͳͲɺ৭ʑͱϋʔυϧ͕͋ͬͨ
    w5FOTPS'MPX
    
    ,FSBTͱ͍ͬͨɺػցֶ श༻ͷϑϨʔϜϫʔΫΛ࢖Θͳ͍ͱݫ͍͠
    w্هͷϑϨʔϜϫʔΫͷ஌͕ࣝෆՄܽ
    wͦ΋ͦ΋ΞϧΰϦζϜͷ஌ࣝ΋ඞཁ
24. $*'"3 IUUQXXXDTUPSPOUPFEVdLSJ[DJGBSIUNM
25. $*'"3 w
    Y
    ͷը૾ຕͷσʔληοτ
    wτϨʔχϯάσʔλ͕ຕ
    wςετσʔλ͕ຕ
    wτϨʔχϯάσʔλͰֶशͨ͠ͷͪɺςε τσʔλΛ࢖ͬͯݕূ͢Δ
26. ৞ΈࠐΈχϡʔϥϧωοτϫʔΫ IUUQTXXXZPVUVCFDPNXBUDI UJNF@DPOUJOVFW2;)$1OXX
27. ৞ΈࠐΈχϡʔϥϧωοτϫʔΫ IUUQTLFSBTJPKB wରԠ͍ͯ͠ΔϥΠϒϥϦͷબఆ
28. wϥΠϒϥϦʹैͬͯίʔσΟϯά model = Sequential() model.add(Conv2D(32, (3, 3), padding='same', input_shape=X_train.shape[1:])) model.add(Activation('relu')) model.add(Conv2D(32, (3, 3))) model.add(Activation('relu')) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Dropout(0.25)) model.add(Conv2D(64, (3, 3), padding='same')) model.add(Activation('relu')) model.add(Conv2D(64, (3, 3))) model.add(Activation('relu')) model.add(MaxPooling2D(pool_size=(2, 2))) model.add(Dropout(0.25)) model.add(Flatten()) model.add(Dense(512)) model.add(Activation('relu')) model.add(Dropout(0.5)) model.add(Dense(nb_classes)) model.add(Activation('softmax')) model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy']) ৞ΈࠐΈχϡʔϥϧωοτϫʔΫ
29. Ͳ͔͜ΒखΛ͚ͭΕ͹͍͍ͷ͔ʜ
30. "84
    SF*OWFOU
    
    /&8
    -"6/$)
    *OUSPEVDJOH
    "NB[PO
    4BHF.BLFS
    .$-
31. IUUQTXXXZPVUVCFDPNXBUDI WQC9ETK;Y@L
32. ֶशϞσϧ࡞੒·Ͱͷγʔέϯε
33. ֶशϞσϧ࡞੒·Ͱͷγʔέϯε
34. ֶशϞσϧ࡞੒·Ͱͷγʔέϯε
35. ֶशϞσϧ࡞੒·Ͱͷγʔέϯε
36. ֶशϞσϧ࡞੒·Ͱͷγʔέϯε
37. ֶशϞσϧ࡞੒·Ͱͷγʔέϯε
38. ֶशϞσϧ࡞੒·Ͱͷγʔέϯε
39. ֶशϞσϧ࡞੒·Ͱͷγʔέϯε 㲔
40. ֶशϞσϧ࡞੒·Ͱͷγʔέϯε 㲔
41. ֶशϞσϧ࡞੒·Ͱͷγʔέϯε
42. OPUFCPPLͷ࡞੒
43. OPUFCPPLͷ࡞੒
44. OPUFCPPLͷ࡞੒
45. OPUFCPPLͷىಈ w*O4FSWJDFʹͳͬͨΒɺ0QFOΛΫϦοΫ͢ Δͱىಈ͢Δ
46. OPUFCPPLͷىಈ IUUQTHJUIVCDPNBXTMBCTBNB[POTBHFNBLFSFYBNQMFTUSFFNBTUFS TBHFNBLFSQZUIPOTELNYOFU@HMVPO@DJGBS wαϯϓϧͷʮTBNQMFOPUFCPPLʯ ʮTBHFNBLFSQZSIPOTELʯ ʮNYOFU@DJGBSʯΛࢼ͠ʹ࣮ߦ͢Δ
47. δϣϒͷ࡞੒ͷ४උ
48. δϣϒͷ࡞੒ͷ४උ wඞཁͳϥΠϒϥϦͷΠϯετʔϧ
49. δϣϒͷ࡞੒ͷ४උ
50. δϣϒͷ࡞੒ w*O
    <
    >
    ͷۭཝΛΫϦοΫ͢Δ
51. δϣϒͷ࡞੒
52. δϣϒͷ࡞੒ wδϣϒ͕࡞੒͞ΕΔͱֶश͕࣮ߦ͞ΕΔ
53. δϣϒͷ࣮ߦ
54. δϣϒͷ࣮ߦ wOPUFCPPL͔ΒֶशͷਐḿΛ֬ೝͰ͖Δ
55. ֶशϞσϧͷ࡞੒ wδϣϒ͕੒ޭ͢ΔͱɺֶशϞσϧ͕࡞੒͞ ΕΔ
56. ֶशϞσϧͷ࡞੒ wδϣϒ͕੒ޭ͢ΔͱɺֶशϞσϧ͕࡞੒͞ ΕΔ
57. ֶशϞσϧͷ࡞੒ wֶशϞσϧͷৄࡉΛ֬ೝ
58. ֶशϞσϧͷ࡞੒
59. ֶशϞσϧͷ࡞੒ wֶशϞσϧ͸μ΢ϯϩʔυ͢Δ͜ͱ͕Մೳ
60. wֶश؀ڥΛ४උ͢Δ·Ͱʹ͔͔͍ͬͯͨί ετΛ௿ݮ
    wֶशͷਐḿΛՄࢹԽ
    wαϯϓϧΛར༻͢Δ͜ͱͰʮࣗ਎ͷֶशʯ ͷͱ͔͔ͬΓʹ͢Δ͜ͱ͕Մೳ ·ͱΊ
61. ࠓޙ΍Γ͍ͨ͜ͱ IUUQTBXTBNB[PODPNKQCMPHTOFXTBNB[POTBHFNBLFS
62. ࠓޙ΍Γ͍ͨ͜ͱ IUUQTBXTBNB[PODPNKQCMPHTOFXTCSJOHNBDIJOFMFBSOJOHUPJPTBQQT VTJOHBQBDIFNYOFUBOEBQQMFDPSFNM
63. ଓ͖͸ϒϩάͰ IUUQTEFWDMBTTNFUIPEKQ
64. ͋Γ͕ͱ͏͍͟͝·ͨ͠