Month: November 2014

H2O and Machine Learning

Working with H2O has been quite an experience so far. Lets look at how to set it up. We can setup H2O as standalone server, install in R or install in Hadoop. Setting it up on standalone is quite simple.

download the zip file
unzip h2o-version.zip
cd into the directory
java -jar h2o.jar

Go to http://localhost:54321/ to see your output. It should look like this:

h2odirect

Installing in R requires slightly complex steps, especially if you are working with Ubuntu or linux.

Install the package by the following command:

install.packages("h2o", repos=(c("http://s3.amazonaws.com/h2orelease/h2o/master/2.8.3.2/R", getOption("repos"))))

Initialize the package and verify that H2O installed properly:

library(h2o)

localH2O = h2o.init()

h2or

Installing in Hadoop requires you to have a Cloudera or Hortonworks or MapR version of Hadoop running on your system because this is what I found inside the h2o/hadoop directory. You can see the drivers for these versions of Hadoop only.

hadoopdriverh2o