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

Watson Studioで かんたん機械学習モデル作成

Kyoko Nishito
December 14, 2018

Watson Studioで かんたん機械学習モデル作成

2018年12月14日オープンソースカンファレンス2018 OSC2018.Enterpriseでの登壇資料です。

Kyoko Nishito

December 14, 2018
Tweet

More Decks by Kyoko Nishito

Other Decks in Technology

Transcript

  1. Watson Studio ) ( Developer Advocate    

    2018 .Enterprise 2018-12-14 () 13:00-13:45 @5F 5A
  2. "  3 • &$ Watson Studio   !"(/'-

    *) #% ., •  Watson Studio Model Builder+ (/ '- *) 
  3. Watson Studio 4 Model Lifecycle Management c c i c

    c i c o d vf AI C a c e st • IBM liu (LocalC r ) • cS IBM C n fI • R, Python, Scala iu •     •    c c i c c i c iu
  4. ) ( Learning ( 5 • Watson Studio DF Fabric

    for Deep Learning (FfDL) • Code Patterns Deploy and use a multi-framework deep learning platform on Kubernetes • Github https://github.com/IBM/FfDL
  5. IBM Developer https://developer.ibm.com/jp/ 6 •    140 

     Code Patterns •    6,000  
  6.  8 dog cat cat dog Positive ID: Pug Unknown

    Unknown Positive ID: Pug dog dog dog dog  
  7.   (1'0+) 9 dog cat cat dog Positive ID:

    Pug Unknown Unknown Positive ID: Pug  ,/'0 Watson Studio  #$" (1'0 Spark ML*&+)  #$%!.- Apache Spark
  8. 3 ) 10 ) 3( ) 3 ) I 3

    ) I 3 i : ) • SEPAL LENGTH: cm • SEPAL WIDTH : cm • PETAL LENGTH: cm • PETAL WIDTH : cm • CLASS: Iris-setosa( ), Iris-versicolour( ) ), Iris-virginica ( )I )
  9. %  11    #!   !

     #  "$% • SEPAL LENGTH:    • SEPAL WIDTH :  #  • PETAL LENGTH: !    • PETAL WIDTH : !  #  • CLASS : Iris-setosa(), Iris-versicolour(), Iris-virginica ()
  10. Watson . 13 IBM Cloud 2. . Watson Studio (

    AI) S/ . 1 Machine Learning ( AI) S/ . 1 Apache Spark ( Web ) S/ . 1 ( ) 
  11. 2. Watson Studio 15 1.   (IBM Cloud )

    2.  Watson Studio   3. Get Started 
  12. 2. Watson Studio  16 4.   5 

    [Next]( [Done])     
  13. 20 5. iris.csv 7 iris_test.csv4 setosa, versicolour, virginica . 10

    4 3 iris_test.csv 4 3 5iris.csv 7  4 
  14. 25 1. [Existing] .4 . 2. Existing Service Instance 1

    Apache Spark 1. [Select] .4 . 
  15. 27 6. [Existing] .4 . 7. Existing Service Instance 1

    Machine Learning 8. [Select] .4 . 
  16. . 32 2. Models Watson Machine Learning models [+ New

    Watson Machine Learning model] 
  17. 6 33 3. [Name] 6 4. Machine Learning Service Machine

    Learning . 5. Select model type: Model builder 6. Select runtime: Apache Spark . 7. Manual 8. [Create] 
  18. . 37 11. . . . . . • Binary

    Classification: (2 . 6 • Multiclass Classification: ( . 6 ) • Regression: ( . ) 3 . . . Multiclass Classification 
  19. . 38 11. . 6. . Train 6 Test 6

    Holdout 6( ) . 
  20. 41 * Model Builder  Binary Classification Multiclass Classification Regression

    Logistic Regression 6 . Y Decision Tree Regression Y Y Y Random Forest Regression Y Y Y Gradient Boosted Tree Regression 6 Y Y Native Bayes Y Linear Regression Y Isotonic Regression . Y 
  21. 56 3. [Predict] iris_test.csv 1 • SEPAL LENGTH : cm

    • SEPAL WIDTH : cm • PETAL LENGTH : . cm • PETAL WIDTH : . cm • Iris-setosa( ), • Iris-versicolour( ) • Iris-virginica ( ) 
  22.  • Watson StudioModel Builder)1(/# % 0.+* • +* #

    %Web"-'  ,   $!&!,  
  23. IBM Cloud!# • %$ '0IBM Cloud!# )3   •

    "*, +&4-5 Watson6 /.API  42 1(  #  https://www.ibm.com/cloud-computing/jp/ja/lite-account/
  24. u R U etr ou kta IBM Partners https://ibm.biz/ibmpartners cphbdkik

    h I E 6 K A (s I E *IIAI E *41 *41 C (m l ,4 ) 5*3 0 K HE I B C – ) 1 3 C B AE 4C HD E HGHAI M A4 https://ibm.biz/buildwith 1,000 IBM Cloud A4 1 12,000l M 132 P S 1 3 ,C K f nid h k T W
  25. IBM@6F •  "Watson 0/' • IBM Cloud$,-  (

    !Zero +TJBot 23#$' • &%  ).*1  
  26. 65