US 12,462,140 B2
System and method for signal conversion in a neural network
Nathan Hackett, Hangzhou (CN)
Assigned to Silergy Semiconductor Technology (Hangzhou) LTD, Hangzhou (CN)
Filed by Silergy Semiconductor Technology (Hangzhou) LTD, Hangzhou (CN)
Filed on May 10, 2021, as Appl. No. 17/315,545.
Claims priority of application No. 202010412693.0 (CN), filed on May 15, 2020.
Prior Publication US 2021/0357724 A1, Nov. 18, 2021
Int. Cl. G06F 17/00 (2019.01); G06N 3/04 (2023.01); H03F 3/45 (2006.01)
CPC G06N 3/04 (2013.01) [H03F 3/45475 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A system for signal conversion in a neural network, the system comprising:
a) a processing module comprising a rectifier circuit having a first diode, a proportional operation circuit, and a current mirror having first and second transistors, and being configured to control an output signal of the processing module to be a fixed value when an input signal of the processing module is in a first interval;
b) the processing module being configured to control the output signal to be in a preset non-linear relationship with the input signal, when the input signal is in a second interval, wherein the processing module is configured to realize the preset non-linear relationship between the output signal and the input signal through exponential calculation;
c) wherein the output signal increases nonlinearly with an increase of the input signal until converging;
d) wherein the processing module comprises a first processing module having the proportional operation circuit and being configured to generate an intermediate signal after rectifying and amplifying the input signal, and a second processing module configured to perform a non-linear processing on the intermediate signal to generate the output signal, wherein the first processing module comprises an operational amplifier coupled to first and second resistors, and the second processing module comprises a second diode connected in series with the first transistor; and
e) wherein the second processing module is configured to obtain a non-linear signal with a natural constant as a base and the intermediate signal as an exponent, and to generate the output signal according to a difference between the preset value and the non-linear signal.