Numenta’s Director of ML Architecture Lawrence Spracklen will be presenting with Nick Ni from Xilinx on optimizing state-of-the-art deep learning networks into sparse topology on July 8th, 2021 at 7:45 AM PDT (4:45 PM CEST).
FPGAs have made a regular evolutional leap forward in terms of new approaches and solutions for both hardware- and software developers. The FPGA Conference Europe addresses that progress across all major manufacturers. It focuses on user-oriented, practically applicable solutions that developers can quickly integrate into their own everyday work. Click here for more information.
In this talk, we will present a neuroscience approach to optimize state-of-the-art deep learning networks into sparse topology and how it can unlock significant performance gains on FPGAs without major loss of accuracy. We will then walk through the FPGA implementation where we exploit the advantage of sparse networks with a unique Domain Specific Architecture (DSA).