An issue here, as with any "big data" or "machine based learning" context is establishing the "noise floor" - or deciding when you have gotten the useful features and are now starting to get into the irrelevant stuff.
-- Raul On Fri, Jun 16, 2017 at 7:14 PM, Skip Cave <[email protected]> wrote: > Code Modernization: Bringing Codes Into the Parallel Age > HPCwire > Doug Black > June 8, 2017 > > Application code must keep up with processor advancements as parallel > computing gains traction, which Intel's Joe Curley says his company is > doing by parallelizing public codes for the newest x86 central-processing > unit generations. In an interview, Curley says the Intel Parallel Computing > Centers produce output that software developers and academics can apply to > teach people cutting-edge code modernization. "We focus efforts on open > source communities and open source codes...to improve the understanding of > how to program in parallel and how to solve problems," Curley says. He > notes the inclusion of artificial intelligence and machine-learning methods > are a recent code modernization development, which when properly > implemented can exponentially boost performance. Curley also says most > projects in this space concentrate on challenging algorithms via > parallelization to "achieve massive increases in performance." Among the > efforts encompassed by Intel's initiative are applications used by > manufacturers in product design, and advanced imaging diagnostics employed > by the medical industry. > > Full Article: https://goo.gl/jdxLuA <https://goo.gl/jdxLuA> > > > Skip Cave > Cave Consulting LLC > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
