US 12,118,463 B1
Weight value decoder of neural network inference circuit
Kenneth Duong, San Jose, CA (US); Jung Ko, San Jose, CA (US); and Steven L. Teig, Menlo Park, CA (US)
Assigned to PERCEIVE CORPORATION, San Jose, CA (US)
Filed by Perceive Corporation, San Jose, CA (US)
Filed on Dec. 14, 2021, as Appl. No. 17/550,177.
Application 17/550,177 is a continuation of application No. 16/457,757, filed on Jun. 28, 2019, granted, now 11,210,586.
Application 16/457,757 is a continuation in part of application No. 16/120,387, filed on Sep. 3, 2018, granted, now 10,740,434, issued on Aug. 11, 2020.
Claims priority of provisional application 62/853,128, filed on May 27, 2019.
Claims priority of provisional application 62/797,910, filed on Jan. 28, 2019.
Claims priority of provisional application 62/792,123, filed on Jan. 14, 2019.
Claims priority of provisional application 62/773,164, filed on Nov. 29, 2018.
Claims priority of provisional application 62/773,162, filed on Nov. 29, 2018.
Claims priority of provisional application 62/753,878, filed on Oct. 31, 2018.
Claims priority of provisional application 62/742,802, filed on Oct. 8, 2018.
Claims priority of provisional application 62/724,589, filed on Aug. 29, 2018.
Claims priority of provisional application 62/660,914, filed on Apr. 20, 2018.
Int. Cl. G06N 3/08 (2023.01); G06N 3/04 (2023.01); G06N 3/065 (2023.01)
CPC G06N 3/08 (2013.01) [G06N 3/04 (2013.01); G06N 3/065 (2023.01)] 19 Claims
OG exemplary drawing
 
1. A neural network inference circuit comprising:
a set of memories for storing encoded weight value data for a plurality of filter portions, the encoded weight value data for each filter portion of a plurality of filter portions comprising:
an identifier for the filter portion;
a first set of data indicating a number of non-zero weight values in the filter portion; and
a second set of data comprising, for each non-zero weight value in the filter portion, additional data regarding the weight; and
a set of computation circuits for decoding the encoded weight value data and using the decoded weight value data to execute a neural network.