Bootstrap1305 beta*
- The Analytics1305 Machine Learning Library using Hadoop
Bootstrap1305 is the scalable version of the Cloud1305 product.
Analyzing huge volumes of data in an exact way requires solving an
optimization problem that has billions of variables. Since most of
the exact solvers have quadratic complexity it is impossible to use
them. Another approach for doing machine learning on huge data volumes
is to use the well known and statistically significant method of
bootstrapping
or bagging.
Map-Reduce is a framework that fits very well in the bootstrap concept. The biggest advantage of the cloud is that you can use a cluster of computers on demand, which would be very expensive to buy and maintain. Moreover setting up hadoop on a cluster is not trivial. Bootstrap1305 is a simple way to run machine learning algorithms using Hadoop. We have setup the AMIs and the scripts so that minimum effort is required by th user in order to run an algorithm on a cluster.
For the moment Support Vector Machine is the only algorithm that has been released under Bootstrap1305 and you may find the documentation here. Moreover, the cost of launching our AMI is the same as that which Amazon charges for it's Elastic MapReduce service. You only pay Amazon charges for data transfers and the tiny signup and monthly fee is to cover Amazon transaction fee.
If you have more questions please contact us at support@analytics1305.com.