![model = est2.fit(est1.fit(
tf2.transform(tf1.transform(data)))
.transform(
tf2.transform(tf1.transform(data)))
)
model = Pipeline(stages=[tf1, tf2, est1, es2]).fit(data)](https://image.slidesharecdn.com/brk3320-180509201013/85/The-Developer-Data-Scientist-Creating-New-Analytics-Driven-Applications-using-Apache-Spark-with-Azure-Databricks-27-320.jpg)
![JFK
IAD
LAX
SFO
SEA
DFW
(b)
(a)
(c)
Search for structural
patterns within a graph.
val paths: DataFrame =
g.find(“(a)-[e1]->(b);
(b)-[e2]->(c);
!(c)-[]->(a)”)
Then filter using vertex
& edge data.
paths.filter(“e1.delay > 20”)](https://image.slidesharecdn.com/brk3320-180509201013/85/The-Developer-Data-Scientist-Creating-New-Analytics-Driven-Applications-using-Apache-Spark-with-Azure-Databricks-61-320.jpg)
The Developer Data Scientist – Creating New Analytics Driven Applications using Apache Spark with Azure Databricks
- 3. Model & ServePrep & Train Databricks HDInsight Data Lake Analytics Custom apps Sensors and devices Store Blobs Data Lake Ingest Data Factory (Data movement, pipelines & orchestration) Machine Learning Cosmos DB SQL Data Warehouse Analysis Services Event Hub IoT Hub SQL Database Analytical dashboards Predictive apps Operational reports Intelligence B I G D ATA & A D VA N C E D A N A LY T I C S AT A G L A N C E Business apps 10 01 SQLKafka
- 4. A fast, easy and collaborative Apache® Spark™ based analytics platform optimized for Azure Best of Databricks Best of Microsoft Designed in collaboration with the founders of Apache Spark One-click set up; streamlined workflows Interactive workspace that enables collaboration between data scientists, data engineers, and business analysts. Native integration with Azure services (Power BI, SQL DW, Cosmos DB, Blob Storage) Enterprise-grade Azure security (Active Directory integration, compliance, enterprise -grade SLAs)
- 5. A Z U R E D A T A B R I C K S Microsoft Azure
- 9. Infrastructure management Data exploration and visualization at scale Time to value - From model iterations to intelligence Integrating with various ML tools to stitch a solution together Operationalize ML models to integrate them into applications
- 10. Optimized Databricks Runtime Engine DATABRICKS I/O SERVERLESS Collaborative Workspace Cloud storage Data warehouses Hadoop storage IoT / streaming data Rest APIs Machine learning models BI tools Data exports Data warehouses Azure Databricks Enhance Productivity Deploy Production Jobs & Workflows APACHE SPARK MULTI-STAGE PIPELINES DATA ENGINEER JOB SCHEDULER NOTIFICATION & LOGS DATA SCIENTIST BUSINESS ANALYST Build on secure & trusted cloud Scale without limits A Z U R E D A T A B R I C K S
- 11. Easy to create and manage compute clusters that auto-scale Rapid development using the integrated workspace that facilitates cross-team collaboration Interactive exploration with notebooks and dashboards Seamless integration with ML eco-system libraries and tools Deep Learning support with GPUs (coming soon in next release)
- 17. Spark SparkSQL Streaming MLlib GraphX
- 19.
- 22. Train model 1 Evaluate Datasource 1 Datasource 2 Datasource 2 Extract featuresExtract features Feature transform 1 Feature transform 2 Feature transform 3 Train model 2 Ensemble
- 23. Simple construction, tuning, and testing for ML workflows
- 27. model = est2.fit(est1.fit( tf2.transform(tf1.transform(data))) .transform( tf2.transform(tf1.transform(data))) ) model = Pipeline(stages=[tf1, tf2, est1, es2]).fit(data)
- 29. 29 Cross Validation ... Best Model Model #1 Training Model #2 Training Feature Extraction Model #3 Training
- 30. Microsoft Confidential Advanced Analytics: Pipeline
- 33. Data Science Software Engineering Prototype (Python/R) Create model Re-implement model for production (Java) Deploy model 3
- 34. Data Science Software Engineering Prototype (Python/R) Create Pipeline • Extract raw features • Transform features • Select key features • Fit multiple models • Combine results to make prediction • Extra implementation work • Different code paths • Synchronization overhead Re-implement Pipeline for production (Java) Deploy Pipeline 3
- 35. Data Science Software Engineering Prototype (Python/R) Create Pipeline Persist model or Pipeline: model.save(“path://...”) Load Pipeline (Scala/Java) Model.load(“path://…”) Deploy in production
- 41. Classification Logistic regression w/ elastic net Naive Bayes Streaming logistic regression Linear SVMs Decision trees Random forests Gradient-boosted trees Multilayer perceptron One-vs-rest Regression Least squares w/ elastic net Isotonic regression Decision trees Random forests Gradient-boosted trees Streaming linear methods Recommendation Alternating Least Squares Frequent itemsets FP-growth Prefix span Clustering • Gaussian mixture models • K-Means • Streaming K-Means • Latent Dirichlet Allocation • Power Iteration Clustering Statistics • Pearson correlation • Spearman correlation • Online summarization • Chi-squared test • Kernel density estimation Linear algebra • Local dense & sparse vectors & matrices • Distributed matrices • Block-partitioned matrix • Row matrix • Indexed row matrix • Coordinate matrix • Matrix decompositions Model import/export Pipelines Feature extraction & selection • Binarizer • Bucketizer • Chi-Squared selection • CountVectorizer • Discrete cosine transform • ElementwiseProduct • Hashing term frequency • Inverse document frequency • MinMaxScaler • Ngram • Normalizer • One-Hot Encoder • PCA • PolynomialExpansion • RFormula • SQLTransformer • Standard scaler • StopWordsRemover • StringIndexer • Tokenizer • StringIndexer • VectorAssembler • VectorIndexer • VectorSlicer • Word2Vec And more… 4
- 42. • Classification • Regression • Recommendation • Clustering • Frequent itemsets 4 • Model import/export • Pipelines • DataFrames • Cross validation • Feature extraction & selection • Statistics • Linear algebra
- 48. Use Azure Databricks for scaling out ML task Leverage well-known model architectures MLLib Pipeline API simplifies ML workflows Leverage pre-trained models for common tasks
- 54. 5 JFK IAD LAX SFO SEA DFW src dest delay tripid SFO SEA 45 105892 3 id city state SEA Seattle WA vertex (node) edge vertex
- 58. JFK IAD LAX SFO SEA DFW src dest delay tripid SFO SEA 45 105892 3 LAX JFK 52 410022 4 id city state SEA Seattle WA SFO San Francisco CA JFK New York NY vertices DataFrame edges DataFrame vertex
- 60. es)
- 61. JFK IAD LAX SFO SEA DFW (b) (a) (c) Search for structural patterns within a graph. val paths: DataFrame = g.find(“(a)-[e1]->(b); (b)-[e2]->(c); !(c)-[]->(a)”) Then filter using vertex & edge data. paths.filter(“e1.delay > 20”)
- 63. Save & load the DataFrames. vertices = sqlContext.read.parquet(...) edges = sqlContext.read.parquet(...) g = GraphFrame(vertices, edges) g.vertices.write.parquet(...) g.edges.write.parquet(...)
Editor's Notes
- #18: Contributions estimated from github commit logs, with some effort to de-duplicate entities.
- #24: No time to mention: User-defined functions (UDFs) Optimizations: code gen, predicate pushdown
- #29: Model training / tuning Regularization: parameter that controls how the linear model does on unseen data There is no single good value for the regularization parameter. One common method to find on is to try out different values. This technique is called CV: you split your training data into 2 sets: one set used to learn some parameters with a given regularization parameter, and another set to evaluate how well we are doing with the given parameter.
- #31: 30
- #36: Note this is loading into Spark.
- #55: 54
- #56: 55
- #57: 56
- #58: 57