An iterative computing framework on both Hadoop MapReduce and Hadoop YARN.
Please visit guagua site for tutorials.
Guagua, a sub-project of Shifu, is a distributed, pluggable and scalable iterative computing framework based on Hadoop MapReduce and YARN.
This graph shows the iterative computing process for Guagua.
Typical use cases for Guagua are distributed machine learning model traing based on Hadoop. By using Guagua, we implement distributed neural network algorithm which can reduce model training time from days to hours on 500GB data sets. For distributed neural network algorithm, it is based on Encog. For any details please check our example source code.
Copyright 2013-2014, eBay Software Foundation under the Apache License.