Handwriting Recognition Using a Cloud Runtime

Handwriting recognition can be a difficult task – while there are many commercial products in place which can be trained for individual users, these programs often have problems abstracting patterns learned in order to work for multiple users. In this work we use the Optical Recognition of Handwritten Digits [3] dataset from the UCI Machine Learning Repository, which contains handwritten digits from 43 people. Using basic neural network code written in R, we analyze different approaches to classification in a distributed environment: Through R itself with Snowfall [7], and with Granules [8, 9] using JRI [1] as well as a binary bridge for communication to the same base R code.