Attending this event?
View analytic
Thursday, September 27 • 15:50 - 16:20
Phylanx – A Distributed Asynchronous Array Processing Toolkit

Sign up or log in to save this to your schedule and see who's attending!

Today, machine learning is being applied in almost all fields of science and industry. We present a new open source project ‘Phylanx’ that opens up opportunities to tap into distributed machine learning on problems where performance is a driving concern. Currently, machine learning applications are usually constrained in their ability to access high performance resources due to the lack of access to efficient parallel runtime systems. We use HPX, the C++ Standard Library for Parallelism and Concurrency, which enables linear algebra algorithms to be optimized for the target platform through cost functions that select data tiling and resource parameters. Phylanx combines highly efficient execution provided by a modular C++ backend with a highly productive Python based front end that many domain scientists are familiar with.

Phylanx offers compelling advances in three areas. First, it delivers scalable machine learning capabilities through its high-performance HPX-based implementations of core distributed parallel algorithms for machine learning frameworks. HPX's efficient support for executing asynchronous dynamic distributed task graphs is a unique technology that makes this possible. Second, it provides exceptional development productivity through support of high-level Python programming interfaces and semantics such that machine learning frameworks and applications easily run in a Phylanx environment, independent of where Phylanx is installed. The portability of HPX to different platforms and its ability to achieve high performance execution is again key. Third, Phylanx provides an avenue to realize high performance machine learning solutions on cloud resources, thereby making it possible to realize "domain science gateways" as a service. Here users build their domain solutions with rich client interface and edge computing tools, with automatic access to Phylanx-optimized algorithms.
This talk will demonstrate modern C++ techniques adopted to the design and implementation of Phylanx as well as demonstrate some exemplar use cases of this new and exciting technology.


Hartmut Kaiser

STE||AR Group, Center for Computation and Technology
Hartmut is a member of the faculty at the CS department at Louisiana State University (LSU) and a senior research scientist at LSU's Center for Computation and Technology (CCT). He received his doctorate from the Technical University of Chemnitz (Germany) in 1988. He is probably best... Read More →

Thursday September 27, 2018 15:50 - 16:20

Attendees (12)