BSSML17 - API and WhizzML
•
1 like•277 views
BSSML17 - Machine Learning Summer School in Curitiba: Rest API and Machine Learning Workflows with WhizzML.
More Related Content
What's hot
What's hot (20)
Similar to BSSML17 - API and WhizzML
Similar to BSSML17 - API and WhizzML (20)
More from BigML, Inc
More from BigML, Inc (20)
Recently uploaded
Recently uploaded (20)
BSSML17 - API and WhizzML
- 1. N O V E M B E R 2 9 , 2 0 1 7
- 2. BigML, Inc 2 API Automating Machine Learning Poul Petersen CIO, BigML, Inc
- 3. BigML, Inc 3API / WhizzML BigML Platform Web-based Frontend Visualizations Distributed Machine Learning Backend SOURCE SERVER DATASET SERVER MODEL SERVER PREDICTION SERVER EVALUATION SERVER SAMPLE SERVER WHIZZML SERVER Tools - https://bigml.com/tools REST API - https://bigml.com/api Smart Infrastructure (auto-deployable, auto-scalable) SERVERS EVENTS GEARMAN QUEUE DESIRED TOPOLOGY AWS COSTS RUNQUEUE SCALER BUSY SCALER AUTO TOPOLOGY AUTO TOPOLOGY AUTO TOPOLOGY AUTO TOPOLOGY ACTUAL TOPOLOGY MESSAGE QUEUE
- 4. BigML, Inc 4API / WhizzML The Need for a ML API • Workflow Automation - reduce drudgery • Abstraction - reuse code • Composability - powerful combinations of APIs • Integration - Dashboard or UI component • Automate deployment • Repeatable results
- 5. BigML, Inc 5API / WhizzML Predictive Applications Collect & Format Data Define ML Problem ETL Model & Evaluate no yes Explore Collect & Format Data Model Automate Consume & Monitor Predict Score Label Drift & Anomaly feature engineer Not Possible tune algorithm Goal Met?
- 6. BigML, Inc 6API / WhizzML BigML API Endpoint https://bigml.io/ / /{id}?{auth} source dataset model ensemble prediction batchprediction evaluation … andromeda dev dev/andromeda • Path elements: • /andromeda specifies the API version (optional) • /dev specifies development mode • if not specified, then latest API in production mode • {id} is required for PUT and DELETE • {auth} contains url parameters username and api_key • api_key can be an alternative key
- 7. BigML, Inc 7API / WhizzML BigML API Endpoint https://bigml.io/...{JSON} {JSON} Operation HTTP Method Semantics CREATE POST Creates a new resource. Returns a JSON document including a unique identifier. RETRIEVE GET Retrieves either a specific resource or a list of resources. UPDATE PUT Updates a resource. Only certain fields are putable. DELETE DELETE Deletes a resource
- 8. BigML, Inc 8API / WhizzML Simple Workflow SOURCE DATASET MODEL PREDICTION
- 9. BigML, Inc 9 API Demo #1
- 10. BigML, Inc 10API / WhizzML BigML Bindings https://github.com/bigmlcom/io
- 11. BigML, Inc 11API / WhizzML Python Binding Overview Operation HTTP Method Binding Method CREATE POST api.create_<resource>(from, {opts}) RETRIEVE GET api.get_<resource>(id, {opts}) api.list_<resource>({opts}) UPDATE PUT api.update_<resource>(id, {opts}) DELETE DELETE api.delete_<resource>(id) • Where <resource> is one of: source, dataset, model, ensemble, evaluation, etc • id is a resource identifier or resource dict • from is a resource identifier, dict, or string depending on context
- 12. BigML, Inc 12 API Demo #2
- 13. BigML, Inc 13API / WhizzML Diabetes Anomalies DIABETES SOURCE DIABETES DATASET TRAIN SET TEST SET ALL MODEL CLEAN DATASET FILTER ALL MODEL ALL EVALUATION CLEAN EVALUATION COMPARE EVALUATIONS ANAOMALY DETECTOR
- 14. BigML, Inc 14 API Demo #3
- 15. BigML, Inc 15API / WhizzML API / Bindings Links • Full API Documentation • https://bigml.com/api • API Endpoint • https://bigml.io • Bindings Index • https://github.com/bigmlcom/io • Python Bindings • https://bigml.readthedocs.io/en/latest/
- 16. BigML, Inc 16 WhizzML Server-Side Machine Learning Automation Poul Petersen CIO, BigML, Inc
- 17. BigML, Inc 17API / WhizzML WhizzML • Complete programming language • Machine Learning operations are first-class citizens • Server-side execution abstracts infrastructure • API First! - Everything is composable • Shareable A Domain-Specific Language (DSL) for automating Machine Learning workflows.
- 18. BigML, Inc 18API / WhizzML WhizzML vs API WhizzML API / Bindings Executes server-side Zero latency Parallelization built-in Sharing built-in Code agnostic workflows Workflows can be UI integrated Requires local execution Every API call has latency Manual parallelization Manual sharing Code specific workflows Workflows external to UI
- 19. BigML, Inc 19API / WhizzML Simple Workflow SOURCE DATASET MODEL PREDICTION
- 20. BigML, Inc 20 WhizzML Demo #1
- 21. BigML, Inc 21API / WhizzML WhizzML Resources LIBRARY SCRIPT EXECUTION • Reusable building-block • Executable code that defines a workflow • Imports libraries. • Defines inputs that parametrize script. • Defines outputs computed by the script. • Given a specific set in inputs computes outputs.
- 22. BigML, Inc 22API / WhizzML WhizzML define • Constants - (define some-variable some-value) • (define pi 3.14159) • (define url "https://example.com") • Expressions - (define some-variable (expression)) • (define source (create-source {opts})) • Functions - (define (func-name opt1… optn) (expression)) • (define (one-click-dataset source) (create-source source {opts})
- 23. BigML, Inc 23 WhizzML Demo #2
- 24. BigML, Inc 24API / WhizzML WhizzML let • Define multiple variables for an expression. • (let ( pi 3.14159 url "https://example.com") (expression)) • Combine with define to create workflows • (define (one-click-model source) ( let (dataset (create-dataset source {opts} )) (create-model dataset {opts})))
- 25. BigML, Inc 25 WhizzML Demo #3
- 26. BigML, Inc 26API / WhizzML WhizzML resources • Handling asynchronous creation • Implicit: (create-dataset (create-source {opts})) • Explicit: (create-dataset (create-and-wait-source {opts})) • Generic: (wait resource) • Fetching full resource object • (define src-id (create-and-wait-source {opts})) (fetch src-id) • Extracting values from resource maps • (define ds-id (create-and-wait-dataset {opts})) (define ds-name (get (fetch ds-id) "name")
- 27. BigML, Inc 27API / WhizzML WhizzML vs Flatline WhizzML Flatline Concerned with resources Turing complete Optimized for parallelization Concerned with datasets More specific to features Optimized for speed
- 28. BigML, Inc 28API / WhizzML WhizzML flatline • Define Flatline transformations inside WhizzML • (flatline "string") • (flatline "(/ (field "Kg") (field "M2"))" • WhizzML variables can be referenced with {{}} • (let (inc 1) (flatline "(+ (field "age") {{inc}})") • Apply to dataset with new_fields • (create-dataset dataset { "new_fields" [ { "name" (str colA " + " colB) "field" (flatline "( + ( field {{colA}} ) ( field {{colB}} ))") }]}))
- 29. BigML, Inc 29API / WhizzML Redfin Workflow Model Predicts Sale Price Sold Homes Compare List to Prediction
- 30. BigML, Inc 30API / WhizzML Redfin Workflow MODEL FILTERSOLD HOMES BATCH PREDICTION NEW FEATURES DATASET DEALS DATASET FILTERFORSALE HOMES NEW FEATURES
- 31. BigML, Inc 31 WhizzML Demo #4
- 32. BigML, Inc 32API / WhizzML WhizzML Resources LIBRARY CITY 1 SOLD HOMES CITY 1 DEALS DATASET EXECUTION CITY 1 FORSALE HOMES SCRIPT
- 33. BigML, Inc 33API / WhizzML WhizzML Resources LIBRARY CITY 2 SOLD HOMES CITY 2 DEALS DATASET EXECUTION CITY 2 FORSALE HOMES SCRIPT
- 34. BigML, Inc 34API / WhizzML WhizzML FE Worth More Worth Less
- 35. BigML, Inc 35API / WhizzML WhizzML FE LATITUDE LONGITUDE REFERENCE LATITUDE REFERENCE LONGITUDE 44,583 -123,296775 44,5638 -123,2794 44,604414 -123,296129 44,5638 -123,2794 44,600108 -123,29707 44,5638 -123,2794 44,603077 -123,295004 44,5638 -123,2794 44,589587 -123,301154 44,5638 -123,2794 Distance (m) 700 30,4 19,38 37,8 23,39 Flatline!
- 36. BigML, Inc 36API / WhizzML WhizzML FE https://en.wikipedia.org/wiki/Haversine_formula
- 37. BigML, Inc 37API / WhizzML WhizzML FE LIBRARY SCRIPT Haversine
- 38. BigML, Inc 38 WhizzML Demo #5
- 39. BigML, Inc 39API / WhizzML Scriptify • "Reifies" a resource into a WhizzML script. • Rapid prototyping meets automation.
- 40. BigML, Inc 40API / WhizzML WhizzML FE Fix Missing Values in a “Meaningful” Way Filter Zeros Model insulin Predict insulin Select insulin Fixed Dataset Amended Dataset Original Dataset Clean Dataset
- 41. BigML, Inc 41API / WhizzML WhizzML Workflow Types Optimizatio Model or Ensemble Best-First Features SMACdown Algorithms Stacked Generalization Gradient boosting Cross Validation Transformatio Flatline Wrappers Remove Anomalies Domain Application Workflow Repetitive Tasks
- 42. BigML, Inc 42API / WhizzML Best-First Features {F1} CHOOSE BEST S = {Fa} {F2} {F3} {F4} Fn S+{F1} S+{F2} S+{F3} S+{F4} S+{Fn-1} CHOOSE BEST S = {Fa, Fb} S+{F1} S+{F2} S+{F3} S+{F4} S+{Fn-1} CHOOSE BEST S = {Fa, Fb, Fc}
- 43. BigML, Inc 43API / WhizzML Model Selection ENSEMBLE LOGISTIC REGRESSION EVALUATION SOURCE DATASET TRAINING TEST MODEL EVALUATIONEVALUATION CHOOSE
- 44. BigML, Inc 44API / WhizzML Model Tuning ENSEMBLE N=20 EVALUATION SOURCE DATASET TRAINING TEST EVALUATIONEVALUATION ENSEMBLE N=10 ENSEMBLE N=1000 CHOOSE
- 45. BigML, Inc 45API / WhizzML SMACdown • How many models? • How many nodes? • Missing splits or not? • Number of random candidates? • Balance the objective? SMACdown can tell you!
- 46. BigML, Inc 46API / WhizzML Higher Level Algorithms • Stacked Generalization • Boosting • Adaboost • Logitboost • Martingale Boosting • Gradient Boosting • Genetic Algorithms
- 47. BigML, Inc 47API / WhizzML Stacked Generalization ENSEMBLE LOGISTIC REGRESSION SOURCE DATASET MODEL BATCH PREDICTION BATCH PREDICTION BATCH PREDICTION EXTENDED DATASET EXTENDED DATASET EXTENDED DATASET LOGISTIC REGRESSION
- 48. BigML, Inc 48API / WhizzML Why WhizzML • Automation is critical to fulfilling the promise of ML • WhizzML can create workflows that: • Automate repetitive tasks. • Automate model tuning and feature selection. • Combine ML models into more powerful algorithms. • Create shareable and re-usable executions.