Deep Learning with H2O & R

Course Overview

Gain firsthand expertise on installing, configuring, and deploying H2O and make use of the R APIs to handle billions of rows of data without sampling and get accurate predictions faster. Through practical guided exercises leverage the built-in machine learning algos such as generalized linear modeling (linear regression, logistic regression, etc.), Naïve Bayes, principal components analysis, time series, k-means clustering, Random Forest, Gradient Boosting, and Deep Learning at scale.

In cloud labs, practice implementing GBM, Random Forest, GLM, GLRM and become familiar with concepts such as Stacking and Super Learning.

Be the expert of deploying complex deep learning models using H2O with R.

At the end of the training, participants will be able to:

  1. Install and configure H2O to work with R, Python, Cloud Providers
  2. Gain a deep understanding of built-in Machine Learning models and usage Access
  3. H2O features through APIs Build and Train multiple models on a single node or in a cluster
  4. Train a generalized linear model, generalized low rank models Install and use H2O
  5. Ensemble to load, train, evaluate model performance Using Storm with H2O for real time prediction
  6. Deliver scalable models that can work on the complex and large datasets

Pre-requisite

Required: Working knowledge of Java, R, Storm, Machine Learning, Deep Learning Models

Duration

3 days

Course Outline

  1. Data Science
  2. H2O
  3. Building a Smarter Application
  4. Combining applications with models
  5. Deploying models into production
  1. Downloading and Unzipping H2O Package
  2. Installing H2O from within R
  3. Installing H2O from within Python
  4. H2O Quickstart with R
  5. H2O Cloud Integration
  1. H2O R Package
  2. Start H2O
  3. Decision Boundaries
  4. Cover Type Dataset
  5. – Exploratory Data Analysis
  6. – Deep Learning Model
  7. – Hyper-Parameter Search
  8. – Checkpointing
  9. – Cross Validation
  • – Model Save & Load
  1. Regression and Binary Classification
  1. Unsupervised Anomaly Detection
  1. Decision Trees
  2. Random Forest
  3. Gradient Boosted Machines
  4. H2O Implementation
  1. Cover Type
  2. Multinomial Model
  3. Binomial Model
  1. Introduction
  2. Basic Model Building(Example)
  3. Plotting Archetypal Features
  4. Imputing Missing Values
  1. Load Training & Test Data
  2. Create Models
  3. Export the best model as POJO
  4. Compile the H2O model as part of the UDF Project
  5. Copy the UDF to the cluster and load into Hive
  6. Score with your UDF
  1. Bagging
  2. Boosting
  3. Stacking / Super Learning
  4. Install H2O Ensemble
  5. Higgs Demo
  6. Start H2O Cluster
  7. Load Data into H2O Cluster
  8. Specify Base Learner & Metalearner
  9. Train an Ensemble
  10. Evaluate Model Performance
  11. Predict
  1. Installing the required software
  2. A brief discussion of the data
  3. Using R to build a gbm model in H2O
  4. Exporting the gbm model as a Java POJO
  5. Copying the generated POJO files into a Storm bolt build environment
  6. Building Storm and the bolt for the model
  7. Running a Storm topology with your model deployed
  8. Watching predictions in real-time

Reviews