US Patent Application 18296165. APPARATUS AND METHOD WITH ACCELERATING ARTIFICIAL NEURAL NETWORK simplified abstract

From WikiPatents
Jump to navigation Jump to search

APPARATUS AND METHOD WITH ACCELERATING ARTIFICIAL NEURAL NETWORK

Inventors

Gopinath Vasanth Mahale of Ankola (IN)


Pramod Parameshwara Udupa of Bengaluru (IN)


Jun-Woo Jang of Suwon-si (KR)


Kiran Kolar Chandrasekharan of Bengaluru (IN)


Sehwan Lee of Suwon-si (KR)


APPARATUS AND METHOD WITH ACCELERATING ARTIFICIAL NEURAL NETWORK - A simplified explanation of the abstract

  • This abstract for appeared for patent application number 18296165 Titled 'APPARATUS AND METHOD WITH ACCELERATING ARTIFICIAL NEURAL NETWORK'

Simplified Explanation

The abstract describes a computer processor apparatus that performs operations on input feature maps (IFMs) using a technique called Winograd convolution (WinConv). The apparatus includes modules for forward transformation, multiplication, accumulation, and inverse transformation. The forward transform module transforms the IFMs, the multiply and accumulate array (MAA) units multiply the transformed IFMs with transformed kernels, and perform an inverse transform operation. Finally, the inverse transform module generates output feature maps (OFMs) based on the result of the first inverse transform operation.


Original Abstract Submitted

A processor-implemented apparatus includes a forward transform module configured to transform input feature maps (IFMs) by performing a forward transform operation in a Winograd convolution (WinConv) domain, multiply and accumulate array (MAA) units configured to multiply the transformed IFMs by transformed kernels and perform a first inverse transform operation based on results of the multiplying, and an inverse transform module configured to generate output feature maps (OFMs) based on a result of the first inverse transform operation.