Apache SystemML AI/ML

1
What Is Apache SystemML ?

• A machine learning system
• Designed to scale to Spark / Hadoop clusters
• Open source / Apache 2 license
• Developed in Java
• Supports R-like and Python-like languages
• Which are designed to scale into the big data
  range
• Automatic optimization at scale for data and
  cluster

2
SystemML Execution Modes

• System ML supports multiple execution modes
• Including
• Standalone
• Spark Batch
• Spark MLContext
• Hadoop Batch
• Java Machine Learning Connector (JMLC)

3
SystemML Dependencies

• System DS forked from ML 1.2
• Current dependencies
• Java 8
• Scala 2.11
• Python 2.7/3.5
• Hadoop 2.6
• Spark 2.1

4
What Is Apache SystemDS ?

• Forked from Apache SystemML 1.2 in September
  2018
• Supports linear algebra programs over matrices
• Replaces the underlying data model and compiler
• Substantially extends the supported
  functionalities
• Supports the whole data science lifecycle
• Data integration, cleaning
• Feature engineering
• Model training
• Over efficient
• Local and distributed ML
• Deployment, serving

5
What Is Apache SystemDS ?

• R-like languages for
• The data-science life cycle stages
• Differing expertise levels
• High-level scripts are compiled into hybrid
  execution plans
• For local, in-memory CPU / GPU operations
• For distributed operations on Apache Spark
• Underlying data model are DataTensors
• Tensors (multi-dimensional arrays) whose first
  dimension
• May have a heterogeneous and nested schema

6
SystemDS Algorithms

• Descriptive Statistics
• Univariate Statistics
• Bivariate Statistics
• Stratified Bivariate Statistics
• Classification
• Multinomial Logistic Regression
• Support Vector Machines
• Binary-Class Support Vector Machines
• Multi-Class Support Vector Machines
• Naive Bayes
• Decision Trees
• Random Forests

7
SystemDS Algorithms

• Clustering
• K-Means Clustering
• Regression
• Linear Regression
• Stepwise Linear Regression
• Generalized Linear Models
• Stepwise Generalized Linear Regression
• Regression Scoring and Prediction
• Matrix Factorization
• Principal Component Analysis
• Matrix Completion via Alternating Minimizations

8
SystemDS Algorithms

• Survival Analysis
• Kaplan-Meier Survival Analysis
• Cox Proportional Hazard Regression Model
• Factorization Machines
• Factorization Machine

9
SystemDS Deep Neural Nets

• Use SystemDS to implement deep neural networks
• Specifying network in Keras format / invoke with
  Keras2DML API
• Specifying network in Caffe format / invoke with
  Caffe2DML API
• Use DML-bodied SystemDS-NN library
• Ease training compute resource issues with
• Native BLAS (Basic Linear Algebra Subprograms)
• SystemDS GPU backend

10
Available Books

• See Big Data Made Easy
• Apress Jan 2015
• See Mastering Apache Spark
• Packt Oct 2015
• See Complete Guide to Open Source Big Data
  Stack
• Apress Jan 2018
• Find the author on Amazon
• www.amazon.com/Michael-Frampton/e/B00NIQDOOM/
• Connect on LinkedIn
• www.linkedin.com/in/mike-frampton-38563020

11
Connect

• Feel free to connect on LinkedIn
• www.linkedin.com/in/mike-frampton-38563020
• See my open source blog at
• open-source-systems.blogspot.com/
• I am always interested in
• New technology
• Opportunities
• Technology based issues
• Big data integration