Oberon Community Platform Forum

Development => General => Topic started by: jabowery on February 16, 2008, 11:46:22 PM



Title: Tensor Decomposition
Post by: jabowery on February 16, 2008, 11:46:22 PM
The addition of matrix math to Oberon for cell processors is great!

This opens up the potential of using Oberon to provide web services for numeric computations -- which means the guys literate in Oberon can use their skills to sell number crunching web services by using special purpose servers loaded with GFLOPS.

As someone who has been following the progress of demand for this kind of service, might I suggest an even more valuable library built atop the cell processors:

Tensor Decomposition

The main value of this is data mining. An example is Peter Turney who started mining large text corpora to create human level performance for SAT analogical reasoning, and has found he is benefiting from "Multislice Projection: Tensor Decomposition (http://www.apperceptual.com/multislice/)".




Title: Re: Tensor Decomposition
Post by: Pat on March 12, 2008, 04:36:26 PM
Is in the making  :)
Pat


Title: Re: Tensor Decomposition
Post by: jabowery on March 13, 2008, 12:04:33 AM
Great news!