Manual

A major challenge facing computer scientists in research and industry today is the automatic analysis and pattern recognition in large sets of data. To address this issue, a new field known as machine learning has developed and produced techniques and algorithms that enable machines to sort, categorize, and make predictions based on provided datasets.

As machine learning tools have matured, so have the sizes of data sets that scientist wish to analyze. To address the issue researchers have turned to solutions provided by the High Performance Computing (HPC) community to distribute the execution of these applications to multiple nodes. These techniques however require users be intimately familiar with HPC in order to effectively make use of distributed resources. In order to address this significant barrier to domain scientist, our team has developed Phylanx, a machine learning platform which exposes a high level Python API but provides HPC level performance.

In this document we describe Phylanx, provide use case examples, and document the components of the system.

• About Phylanx
• Software Stack
  • External Libraries
  • Frontend
  • Execution Tree
  • Instrumentation
  • Visualization Tools
• LRA Example
  • Using Phylanx with the Python Frontend
  • Using Phylanx with PhySL