US 12,260,313 B2
Artificial neural network bypass
Saideep Tiku, Fort Collins, CO (US); and Poorna Kale, Folsom, CA (US)
Assigned to Micron Technology, Inc., Boise, ID (US)
Filed by Micron Technology, Inc., Boise, ID (US)
Filed on Nov. 18, 2020, as Appl. No. 16/951,768.
Prior Publication US 2022/0156559 A1, May 19, 2022
Int. Cl. G06N 3/045 (2023.01); G06N 3/08 (2023.01)
CPC G06N 3/045 (2023.01) [G06N 3/08 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A non-transitory machine-readable medium storing instructions executable to:
execute, at logic of a memory device in a first iteration, a second number of layers of a first artificial neural network to generate a first output probability of a first output and a second output probability of a second output;
execute, at the logic of the memory device that comprises the logic and a memory array, a first number of layers of the first artificial neural network;
determine whether a difference between the first output probability and the second output probability is greater than a threshold;
responsive to executing the first number of layers of the first artificial neural network, responsive to determining that the difference is greater than the threshold, and responsive to determining that the difference is less than a different threshold, execute, at the logic of the memory device in a second iteration, a second number of layers of a second artificial neural network to bypass a third number of layers of the first artificial neural network;
responsive to executing the first number of layers of the first artificial neural network and responsive to determining that the difference is less than the threshold, execute at the logic of the memory device, the third number of layers of the first artificial neural network; and
responsive to executing the first number of layers of the first artificial neural network and responsive to determining that the difference is greater than the different threshold, execute at the logic of the memory device in the second iteration, the second number of layers of the second artificial neural network to bypass the third number of layers of the first artificial neural network and update a confidence table for a fourth number of layers of the second artificial neural network.