US 12,436,738 B2
Method and apparatus with data processing
Seungwon Lee, Hwaseong-si (KR); Namhyung Kim, Seoul (KR); Hanmin Park, Seoul (KR); and Kiyoung Choi, Seoul (KR)
Assigned to Samsung Electronics Co., Ltd., Suwon-si (KR); and Seoul National University R&DB Foundation, Seoul (KR)
Filed by Samsung Electronics Co., Ltd., Suwon-si (KR); and Seoul National University R&DB Foundation, Seoul (KR)
Filed on Mar. 8, 2022, as Appl. No. 17/689,454.
Application 17/689,454 is a continuation of application No. 16/843,199, filed on Apr. 8, 2020, granted, now 11,301,209.
Claims priority of application No. 10-2019-0061458 (KR), filed on May 24, 2019.
Prior Publication US 2022/0188070 A1, Jun. 16, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 7/02 (2006.01); G06F 7/544 (2006.01); G06F 7/74 (2006.01); G06N 3/02 (2006.01)
CPC G06F 7/02 (2013.01) [G06F 7/544 (2013.01); G06F 7/74 (2013.01); G06N 3/02 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A processor-implemented method of processing neural network data, the method comprising:
setting first limit data by performing a first operation based on first input data and weight data generated from weights included in a filter;
comparing the first limit data with an intermediate result of a second operation performed based on second input data and the weight data; and
determining whether to perform a subsequent second operation based on a result of the comparing,
wherein the weight data is generated based on values of the encoded weights corresponding to same bit positions of the encoded weights, and
the encoded weights are weights encoded based on an inverted two's complement representation format of the weights.