Invention Grant
- Patent Title: Optimization of sparsified neural network layers for semi-digital crossbar architectures
-
Application No.: US16849638Application Date: 2020-04-15
-
Publication No.: US11556768B2Publication Date: 2023-01-17
- Inventor: Ryan Hatcher , Titash Rakshit , Jorge Kittl , Dharmendar Palle , Joon Goo Hong
- Applicant: Samsung Electronics Co., Ltd.
- Applicant Address: KR Gyeonggi-do
- Assignee: Samsung Electronics Co., Ltd.
- Current Assignee: Samsung Electronics Co., Ltd.
- Current Assignee Address: KR Gyeonggi-do
- Agency: The Farrell Law Firm, P.C.
- Main IPC: G06N3/063
- IPC: G06N3/063 ; G06N3/02

Abstract:
A method and system are provided. The method includes mapping a binary matrix to an undirected graph form, applying a two-way graph partition algorithm to the mapped binary matrix that minimizes edge cuts between partitions in the mapped binary matrix, applying a greedy algorithm recursively to find a set of row or column permutations that maximizes a transfer of non-zeros from sparse blocks to nonsparse blocks, and sparsifying or densifying the binary matrix according to the applied greedy algorithm.
Public/Granted literature
- US20210133544A1 OPTIMIZATION OF SPARSIFIED NEURAL NETWORK LAYERS FOR SEMI-DIGITAL CROSSBAR ARCHITECTURES Public/Granted day:2021-05-06
Information query