IP Library Granted Patent US 12,248,367
Granted Patent B2
US 12,248,367 · App. 17/035,795 · Granted Mar 11, 2025

Software defined redundant allocation safety mechanism in an artificial neural network processor

Inventors: Avi Baum (Givat Shmuel, IL); Daniel Chibotero (Kirat Ono, IL); Roi Seznayov (Raanana, IL); Or Danon (Kirat Ono, IL); Ori Katz (Tel-Aviv, IL); Guy Kaminitz (Kfar Saba, IL)
G06F11/1004G06F11/27G06N3/02
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,248,367
App. No.
17/035,795
Granted
Mar 11, 2025
Kind
B2
Abstract

Novel and useful system and methods of several functional safety mechanisms for use in an artificial neural network (ANN) processor. The mechanisms can be deployed individually or in combination to provide a desired level of safety in neural networks. Multiple strategies are applied involving redundancy by design, redundancy through spatial mapping as well as self-tuning procedures that modify static (weights) and monitor dynamic (activations) behavior. The various mechanisms of the present invention address ANN system level safety in situ, as a system level strategy that is tightly coupled with the processor architecture. The NN processor incorporates several functional safety concepts which reduce its risk of failure that occurs during operation from going unnoticed. The mechanisms function to detect and promptly flag and report the occurrence of an error with some mechanisms capable of correction as well. The safety mechanisms cover data stream fault detection, software defined redundant allocation, cluster interlayer safety, cluster intralayer safety, layer control unit (LCU) instruction addressing, weights storage safety, and neural network intermediate results storage safety.

Claims (36)

1. A method of end to end failure detection for use in a neural network (NN) processor configured to implement a target neural network, the method comprising:

providing a plurality of unallocated redundant hardware resources in said neural network processor, said unallocated redundant hardware resources including compute, memory, routing, and control elements;

allocating during an offline compilation process via a compiler a main computational path from said plurality of unallocated redundant hardware resources, said main computational path to be protected from end to end failures;

allocating during said offline compilation process via said compiler one or more redundant computational paths from said plurality of unallocated redundant hardware resources different from said plurality of unallocated redundant hardware resources allocated to said main computational path, where said compiler ensures that said main computational path and said one or more redundant computational paths go through and use different resources with no overlap of compute, memory, routing, and control hardware resource elements, wherein said main computational path and said one or more redundant computational paths each function to perform the same calculations for the same target neural network, said one or more redundant computational paths operative to protect said main computational path from end to end failures whereby whole or partial layers of a whole or partial network are allocated;

providing the same input data to said main computational path and said one or more redundant computational paths at the same time;

calculating cyclic redundancy code (CRC) checksums on tensor stream data output from said main computational path and said one or more redundant computational paths;

comparing said CRC checksums from said main computational path and said one or more redundant computational paths with each other;

detecting an error if said calculated CRC checksums do not match; and

wherein said main computational path and said one or more redundant computational paths use different data resources selected from a group consisting of stream managers (SMs), portions of L4 memory allocated to said stream managers, input buffers (IBs), portions of L3 memory, input aligners (IAs), subclusters (SCs), activation processing units (APUs), and output buffers (OBs).

2. The method according to claim 1 , wherein said allocation is determined a priori by a compiler in accordance with desired performance goals for a target neural network.

3. The method according to claim 1 , wherein said main computational path and said one or more redundant computational paths are functionally identical to each other.

4. The method according to claim 1 , wherein said main computational path and said one or more redundant computational paths use different control resources selected from a group consisting of layer controller units (LCUs) and memory management units (MMUs).

5. The method according to claim 1 , wherein hardware resources for said main computational path and said one or more redundant computational paths are allocated in different clusters for each path.

6. The method according to claim 1 , wherein path separation begins within said NN processor at one or more input buffers (IBs) or L3 memory.

7. The method according to claim 1 , further comprising providing built in self-test (BIST) ability where one of said calculated CRC checksums is intentionally altered so as to force generation of an error signal.

8. The method according to claim 1 , wherein hardware resources for said main computational path and said one or more redundant computational paths are allocated on a per layer basis.

9. An apparatus for end to end failure detection for use in a neural network (NN) processor configured to implement a target neural network, comprising:

a plurality of unallocated redundant hardware resources within said neural network processor, said redundant hardware resources including compute, memory, routing, and control elements;

a main computational path allocated during an offline compilation process via a compiler from said plurality of unallocated redundant hardware resources, said main computational path to be protected from end to end failures;

one or more redundant computational paths allocated during said offline compilation process via said compiler from said plurality of unallocated redundant hardware resources different from said plurality of unallocated redundant hardware resources allocated to said main computational path, where said compiler ensures that said main computational path and said one or more redundant computational paths go through and use different resources with no overlap of compute, memory, routing, and control hardware resource elements, wherein said main computational path and said one or more redundant computational paths each function to perform the same calculations for the same target neural network, said one or more redundant computational paths operative to protect said main computational path from end to end failures whereby the same input data is provided to said main computational path and said one or more redundant computational paths at the same time, and whereby whole or partial layers of a whole or partial network are allocated;

one or more cyclic redundancy code (CRC) engines operative to generate CRC checksums on tensor stream data output from said main computational path and said one or more redundant computational paths, compare said CRC checksums from said main computational path and said one or more redundant computational paths with each other, and to detect an error if said calculated CRC checksums do not match; and

wherein said main computational path and said one or more redundant computational paths use different data resources selected from a group consisting of stream managers (SMs), portions of L4 memory allocated to said stream managers, input buffers (IBs), portions of L3 memory, input aligners (IAs), subclusters (SCs), activation processing units (APUs), and output buffers (OBs).

10. The apparatus according to claim 9 , wherein said allocation is determined a priori by a compiler in accordance with desired performance goals for a neural network.

11. The apparatus according to claim 9 , wherein said main computational path and said one or more redundant computational paths are functionally identical to each other.

12. The apparatus according to claim 9 , wherein said main computational path and said one or more redundant computational paths use different control resources selected from a group consisting of layer controller units (LCUs) and memory management units (MMUs).

13. The apparatus according to claim 9 , wherein hardware resources for said main computational path and said one or more redundant computational paths are allocated in different clusters for each path.

14. The apparatus according to claim 9 , wherein path separation begins within said NN processor at one or more input buffers (IBs) or L3 memory.

15. A method of end to end failure detection for use in a neural network processor configured to implement a target neural network, the method comprising:

providing a plurality of unallocated redundant hardware resources in said neural network processor, said unallocated redundant hardware resources including compute, memory, routing, and control elements;

determining a resource allocation scheme in accordance with desired performance goals for a target neural network;

allocating during an offline compilation process via a compiler a plurality of redundant computational tensor data flow paths from said plurality of unallocated redundant hardware resources, where said compiler ensures that said plurality of redundant computational tensor data flow paths go through and use different resources, said plurality of redundant computational tensor data flow paths functionally identical to each other and operative to provide protection from end to end failures by way of said redundancy wherein each redundant computation tensor data flow path uses different unallocated redundant hardware resources with no overlap of compute, memory, routing, and control hardware resource elements, wherein said plurality of redundant computational tensor data flow paths each function to perform the same calculations for the same target neural network, whereby whole or partial layers of a whole or partial network are allocated;

providing the same input data to each of said redundant computational tensor data flow paths at the same time;

comparing cyclic redundancy code (CRC) checksums generated for each redundant computation tensor data flow path and detecting an error if a mismatch is detected; and

wherein said plurality of redundant computational tensor data flow paths use different data resources selected from a group consisting of stream managers (SMs), portions of L4 memory allocated to said stream managers, input buffers (IBs), portions of L3 memory, input aligners (IAs), subclusters (SCs), activation processing units (APUs), and output buffers (OBs).

16. The method according to claim 15 , wherein said plurality of redundant computational tensor data flow paths use different control resources selected from a group consisting of layer controller units (LCUs) and memory management units (MMUs).

17. The method according to claim 15 , wherein hardware resources for said plurality of redundant computational tensor data flow paths are allocated in different clusters for each path.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Sep 29, 2020
From: BAUM, AVI; CHIBOTERO, DANIEL; SEZNAYOV, ROI; DANON, OR; KATZ, ORI; KAMINITZ, GUY
To: HAILO TECHNOLOGIES LTD.
Reel/Frame 053910/0951 →
Continuity (1)
Related Publication 20220100601A1 · Mar 31, 2022
References Cited (360)
US 4450533A · Petit · 1984 [cited by applicant]
US 4943931A · Allen · 1990 [cited by applicant]
US 4979124A · Sachse et al. · 1990 [cited by applicant]
US 4994982A · Duranton et al. · 1991 [cited by applicant]
US 5003490A · Castelaz et al. · 1991 [cited by applicant]
US 5008833A · Agranat et al. · 1991 [cited by applicant]
US 5014235A · Morton · 1991 [cited by applicant]
US 5056037A · Eberhardt · 1991 [cited by applicant]
US 5080464A · Toyoda · 1992 [cited by applicant]
US 5091864A · Baji et al. · 1992 [cited by applicant]
US 5148515A · Vassiliadis et al. · 1992 [cited by applicant]
US 5195169A · Kamiya et al. · 1993 [cited by applicant]
US 5195170A · Eberhardt · 1993 [cited by applicant]
US 5257358A · Cohen · 1993 [cited by applicant]
US 5274743A · Imondi et al. · 1993 [cited by applicant]
US 5278945A · Basehore et al. · 1994 [cited by applicant]
US 5293456A · Guez et al. · 1994 [cited by applicant]
US 5293459A · Duranton et al. · 1994 [cited by applicant]
US 5297232A · Murphy · 1994 [cited by applicant]
US 5365460A · Chung · 1994 [cited by applicant]
US 5367612A · Bozich et al. · 1994 [cited by applicant]
US 5408645A · Ikeda et al. · 1995 [cited by applicant]
US 5423001A · Ueda · 1995 [cited by applicant]
US 5430830A · Frank et al. · 1995 [cited by applicant]
US 5444822A · Shinohara · 1995 [cited by applicant]
US 5448681A · Khan · 1995 [cited by applicant]
US 5450315A · Stefanski · 1995 [cited by applicant]
US 5465204A · Sekine et al. · 1995 [cited by applicant]
US 5479579A · Duong et al. · 1995 [cited by applicant]
US 5485545A · Kojima et al. · 1996 [cited by applicant]
US 5486996A · Samad et al. · 1996 [cited by applicant]
US 5493631A · Huang et al. · 1996 [cited by applicant]
US 5504839A · Mobus · 1996 [cited by applicant]
US 5519811A · Yoneda et al. · 1996 [cited by applicant]
US 5521985A · Camp, Jr. et al. · 1996 [cited by applicant]
US 5533383A · Gtreene et al. · 1996 [cited by applicant]
US 5537513A · DeVille · 1996 [cited by applicant]
US 5583771A · Lynch et al. · 1996 [cited by applicant]
US 5615305A · Nunally · 1997 [cited by applicant]
US 5627943A · Yoneda et al. · 1997 [cited by applicant]
US 5689622A · Higashino et al. · 1997 [cited by applicant]
US 5704016A · Shigematsu et al. · 1997 [cited by applicant]
US 5705956A · Neely · 1998 [cited by applicant]
US 5717832A · Steimle et al. · 1998 [cited by applicant]
US 5740325A · Wang · 1998 [cited by applicant]
US 5745652A · Bigus · 1998 [cited by applicant]
US 5748849A · Gobert · 1998 [cited by applicant]
US 5796925A · Deville · 1998 [cited by applicant]
US 5944841A · Christie · 1999 [cited by applicant]
US 5956703A · Turner et al. · 1999 [cited by applicant]
US 5978782A · Neely · 1999 [cited by applicant]
US 6267298B1 · Campbell · 2001 [cited by applicant]
US 6317658B1 · Vian · 2001 [cited by applicant]
US 6326912B1 · Fujimori · 2001 [cited by applicant]
US 6366236B1 · Farmer et al. · 2002 [cited by applicant]
US 6421359B1 · Bennett et al. · 2002 [cited by applicant]
US 6434541B1 · Tawel et al. · 2002 [cited by applicant]
US 6446238B1 · Canestaro et al. · 2002 [cited by applicant]
US 6493689B2 · Kotoulas et al. · 2002 [cited by applicant]
US 6523018B1 · Louis et al. · 2003 [cited by applicant]
US 6529614B1 · Chao et al. · 2003 [cited by applicant]
US 6539368B1 · Chernikov et al. · 2003 [cited by applicant]
US 6542985B1 · Johnson et al. · 2003 [cited by applicant]
US 6567953B1 · Pomerantz · 2003 [cited by applicant]
US 6578020B1 · Nguyen · 2003 [cited by applicant]
US 6622125B1 · Cragun et al. · 2003 [cited by applicant]
US 6629089B1 · Supino · 2003 [cited by applicant]
US 6665813B1 · Forsman et al. · 2003 [cited by applicant]
US 6671832B1 · Apisdorf · 2003 [cited by applicant]
US 6751529B1 · Fouche · 2004 [cited by applicant]
US 6836767B2 · McBride · 2004 [cited by applicant]
US 6882992B1 · Werbos · 2005 [cited by applicant]
US 7039233B2 · Mori · 2006 [cited by applicant]
US 7069256B1 · Campos · 2006 [cited by applicant]
US 7082419B1 · Lightowler · 2006 [cited by applicant]
US 7113970B2 · Chiueh · 2006 [cited by applicant]
US 7117045B2 · Hittle et al. · 2006 [cited by applicant]
US 7174325B1 · Ascoli · 2007 [cited by applicant]
US 7302619B1 · Tompkins et al. · 2007 [cited by applicant]
US 7308322B1 · Discenzo et al. · 2007 [cited by applicant]
US 7647284B2 · Prokhorov · 2010 [cited by applicant]
US 8081816B1 · Irick et al. · 2011 [cited by applicant]
US 8126828B2 · Snook et al. · 2012 [cited by applicant]
US 8131659B2 · Xu et al. · 2012 [cited by applicant]
US 8145983B1 · Chaichanavong · 2012 [cited by applicant]
US 8321769B1 · Yeo · 2012 [cited by applicant]
US 8326782B2 · Snook et al. · 2012 [cited by applicant]
US 8391306B2 · Ito et al. · 2013 [cited by applicant]
US 8745466B2 · Kaynak · 2014 [cited by applicant]
US 8892485B2 · Aparin et al. · 2014 [cited by applicant]
US 8909576B2 · Akopyan · 2014 [cited by applicant]
US 8996350B1 · Dub · 2015 [cited by applicant]
US 8996967B2 · Kaynak · 2015 [cited by applicant]
US 9014181B2 · Lakshman et al. · 2015 [cited by applicant]
US 9077491B2 · Mar et al. · 2015 [cited by applicant]
US 9129220B2 · Aparin et al. · 2015 [cited by applicant]
US 9129222B2 · Aparin · 2015 [cited by applicant]
US 9153230B2 · Maaninen · 2015 [cited by applicant]
US 9159020B2 · Alcarez-Icaza Rivera et al. · 2015 [cited by applicant]
US 9195656B2 · Fructuso et al. · 2015 [cited by applicant]
US 9306965B1 · Grossman et al. · 2016 [cited by applicant]
US 9361104B2 · Morrison · 2016 [cited by examiner]
US 9400954B2 · Modha · 2016 [cited by applicant]
US 9466022B2 · Alcarez-Icaza Rivera et al. · 2016 [cited by applicant]
US 9477925B2 · Seide et al. · 2016 [cited by applicant]
US 9530092B2 · Chatterjee et al. · 2016 [cited by applicant]
US 9542642B2 · Wood · 2017 [cited by applicant]
US 9542643B2 · Levin et al. · 2017 [cited by applicant]
US 9558444B2 · Palmer et al. · 2017 [cited by applicant]
US 9607217B2 · Cetintas et al. · 2017 [cited by applicant]
US 10019470B2 · Birdwell et al. · 2018 [cited by applicant]
US 10073816B1 · Lu et al. · 2018 [cited by applicant]
US 10339041B2 · Rangan et al. · 2019 [cited by applicant]
US 10417089B2 · Oboukhov · 2019 [cited by applicant]
US 10430706B2 · Henry et al. · 2019 [cited by applicant]
US 10699189B2 · Lie et al. · 2020 [cited by applicant]
US 10789734B2 · Pan et al. · 2020 [cited by applicant]
US 10936569B1 · Baskaran et al. · 2021 [cited by applicant]
US 11232016B1 · Huynh · 2022 [cited by examiner]
US 11392829B1 · Pool · 2022 [cited by applicant]
US 20020112204A1 · Biskup et al. · 2002 [cited by applicant]
US 20020161813A1 · Chiueh · 2002 [cited by applicant]
US 20020184556A1 · Hashemi · 2002 [cited by applicant]
US 20030069516A1 · Becker et al. · 2003 [cited by applicant]
US 20030115537A1 · Boyer et al. · 2003 [cited by applicant]
US 20040167418A1 · Nguyen et al. · 2004 [cited by applicant]
US 20050193357A1 · Honary et al. · 2005 [cited by applicant]
US 20050256420A1 · Noman et al. · 2005 [cited by applicant]
US 20060015703A1 · Ramchandran · 2006 [cited by applicant]
US 20060085558A1 · Solomon · 2006 [cited by applicant]
US 20060167784A1 · Hoffberg · 2006 [cited by applicant]
US 20070266277A1 · Oyamada · 2007 [cited by applicant]
US 20080117220A1 · Gorchetchnikov et al. · 2008 [cited by applicant]
US 20090276666A1 · Haley et al. · 2009 [cited by applicant]
US 20100100514A1 · Raymond et al. · 2010 [cited by applicant]
US 20100205399A1 · Bean et al. · 2010 [cited by applicant]
US 20100235686A1 · Sato · 2010 [cited by applicant]
US 20100313104A1 · Bommena et al. · 2010 [cited by applicant]
US 20100317420A1 · Hoffberg · 2010 [cited by applicant]
US 20110041013A1 · Ingimundarson · 2011 [cited by applicant]
US 20120179898A1 · Betouin et al. · 2012 [cited by applicant]
US 20130079842A1 · Mokelke et al. · 2013 [cited by applicant]
US 20130254892A1 · Kaufman · 2013 [cited by applicant]
US 20140032457A1 · Palmer et al. · 2014 [cited by applicant]
US 20140156576A1 · Nugent · 2014 [cited by applicant]
US 20140344203A1 · Ahn · 2014 [cited by applicant]
US 20150026104A1 · Tambos · 2015 [cited by applicant]
US 20150081753A1 · Srinivasan · 2015 [cited by applicant]
US 20150106311A1 · Birdwell et al. · 2015 [cited by applicant]
US 20150106317A1 · Rangan et al. · 2015 [cited by applicant]
US 20150134580A1 · Wilson · 2015 [cited by applicant]
US 20150170021A1 · Lupon et al. · 2015 [cited by applicant]
US 20150212861A1 · Canoy et al. · 2015 [cited by applicant]
US 20150269480A1 · Levin et al. · 2015 [cited by applicant]
US 20150324690A1 · Chilimbi et al. · 2015 [cited by applicant]
US 20150324691A1 · Dropps et al. · 2015 [cited by applicant]
US 20150324692A1 · Ritchey et al. · 2015 [cited by applicant]
US 20160086077A1 · Akopyan et al. · 2016 [cited by applicant]
US 20160179434A1 · Herrero et al. · 2016 [cited by applicant]
US 20160210550A1 · Merrill · 2016 [cited by applicant]
US 20160232442A1 · Fan et al. · 2016 [cited by applicant]
US 20160275398A1 · Corvinelli et al. · 2016 [cited by applicant]
US 20160321537A1 · Akopyan et al. · 2016 [cited by applicant]
US 20160328646A1 · Lin · 2016 [cited by applicant]
US 20160335119A1 · Merrill et al. · 2016 [cited by applicant]
US 20160335534A1 · Nere et al. · 2016 [cited by applicant]
US 20170011288A1 · Brothers · 2017 [cited by applicant]
US 20170017876A1 · Modha · 2017 [cited by applicant]
US 20170024644A1 · Van Der Made et al. · 2017 [cited by applicant]
US 20170083469A1 · Catthoor · 2017 [cited by applicant]
US 20170090929A1 · Muttik et al. · 2017 [cited by applicant]
US 20170091614A1 · Amir et al. · 2017 [cited by applicant]
US 20170093526A1 · Mula et al. · 2017 [cited by applicant]
US 20170103313A1 · Ross et al. · 2017 [cited by applicant]
US 20170110093A1 · Stein et al. · 2017 [cited by applicant]
US 20170147301A1 · Rong · 2017 [cited by applicant]
US 20170206437A1 · Hachiya et al. · 2017 [cited by applicant]
US 20170228638A1 · Danihelka · 2017 [cited by applicant]
US 20170277658A1 · Pratas · 2017 [cited by applicant]
US 20170330509A1 · Cok et al. · 2017 [cited by applicant]
US 20170345181A1 · Yu et al. · 2017 [cited by applicant]
US 20180018558A1 · Lee et al. · 2018 [cited by applicant]
US 20180046894A1 · Yao · 2018 [cited by applicant]
US 20180139153A1 · Moradi · 2018 [cited by applicant]
US 20180189231A1 · Fleming · 2018 [cited by applicant]
US 20180189631A1 · Sumbul · 2018 [cited by applicant]
US 20180232617A1 · Rozen et al. · 2018 [cited by applicant]
US 20180260220A1 · Lacy et al. · 2018 [cited by applicant]
US 20180322382A1 · Mellempudi et al. · 2018 [cited by applicant]
US 20180336462A1 · Brothers · 2018 [cited by applicant]
US 20180373981A1 · Hu · 2018 [cited by examiner]
US 20180375729A1 · Tzoreff et al. · 2018 [cited by applicant]
US 20190004800A1 · Ong · 2019 [cited by applicant]
US 20190079729A1 · Aimone · 2019 [cited by applicant]
US 20190158223A1 · Carlough et al. · 2019 [cited by applicant]
US 20190197350A1 · Park · 2019 [cited by applicant]
US 20190373588A1 · Bae · 2019 [cited by applicant]
US 20190392296A1 · Brady · 2019 [cited by examiner]
US 20200026992A1 · Zhang · 2020 [cited by applicant]
US 20200104717A1 · Alistarh · 2020 [cited by applicant]
US 20200136943A1 · Banyai · 2020 [cited by examiner]
US 20200154482A1 · Ahn · 2020 [cited by applicant]
US 20200174829A1 · Shao · 2020 [cited by examiner]
US 20200184321A1 · Fulop · 2020 [cited by examiner]
US 20200193297A1 · Verhoef · 2020 [cited by applicant]
US 20200202198A1 · Lee et al. · 2020 [cited by applicant]
US 20200024496A1 · Bhorkar · 2020 [cited by applicant]
US 20200285949A1 · Baum · 2020 [cited by applicant]
US 20200285950A1 · Baum et al. · 2020 [cited by applicant]
US 20200285997A1 · Bhattacharyya · 2020 [cited by applicant]
US 20200322135A1 · Kupwade Patil · 2020 [cited by applicant]
US 20200379740A1 · Paek · 2020 [cited by examiner]
US 20200379841A1 · Cassidy et al. · 2020 [cited by applicant]
US 20200380369A1 · Case · 2020 [cited by applicant]
US 20200382236A1 · Lee · 2020 [cited by examiner]
US 20210053574A1 · Bielby et al. · 2021 [cited by applicant]
US 20210073127A1 · Bielby et al. · 2021 [cited by applicant]
US 20210132678A1 · An · 2021 [cited by applicant]
US 20210158169A1 · Kim · 2021 [cited by applicant]
US 20210256384A1 · Wang · 2021 [cited by applicant]
US 20210279055A1 · Saxena · 2021 [cited by examiner]
US 20220108157A1 · Hunter · 2022 [cited by applicant]
CA 2019099 · 1990 [cited by applicant]
CA 2002681A1 · 1991 [cited by applicant]
CA 2383318A1 · 2003 [cited by applicant]
CA 2927171A1 · 2016 [cited by applicant]
CN 101452258 · 2009 [cited by applicant]
CN 101493677 · 2009 [cited by applicant]
CN 101809597 · 2010 [cited by applicant]
CN 201927073 · 2011 [cited by applicant]
CN 101882238 · 2012 [cited by applicant]
CN 102490120 · 2012 [cited by applicant]
CN 103778468 · 2014 [cited by applicant]
CN 104107507 · 2014 [cited by applicant]
CN 104459064 · 2015 [cited by applicant]
CN 104493826 · 2015 [cited by applicant]
CN 104899641 · 2015 [cited by applicant]
CN 105184366A · 2015 [cited by applicant]
CN 105260776 · 2016 [cited by applicant]
CN 105311750 · 2016 [cited by applicant]
CN 105653790 · 2016 [cited by applicant]
CN 105676649 · 2016 [cited by applicant]
CN 105678379 · 2016 [cited by applicant]
CN 105844330 · 2016 [cited by applicant]
CN 105930902 · 2016 [cited by applicant]
CN 106022468 · 2016 [cited by applicant]
CN 106355246 · 2017 [cited by applicant]
CN 106447034 · 2017 [cited by applicant]
CN 106503796 · 2017 [cited by applicant]
CN 106529670 · 2017 [cited by applicant]
DE 4310279A1 · 1993 [cited by applicant]
DE 19718224 · 1997 [cited by applicant]
EP 0370543B1 · 1990 [cited by applicant]
EP 0388806A2 · 1990 [cited by applicant]
EP 0411761 · 1991 [cited by applicant]
EP 0465241B1 · 1992 [cited by applicant]
EP 525543B1 · 1993 [cited by applicant]
EP 568146B1 · 1993 [cited by applicant]
EP 575716A1 · 1993 [cited by applicant]
EP 2825974 · 2015 [cited by applicant]
EP 3035249 · 2016 [cited by applicant]
EP 3276540 · 2018 [cited by applicant]
FR 2724029B1 · 1997 [cited by applicant]
FR 2780919 · 2000 [cited by applicant]
GB 2245401 · 1992 [cited by applicant]
GB 2351885 · 2001 [cited by applicant]
JP H02157832 · 1990 [cited by applicant]
JP H02206867 · 1990 [cited by applicant]
JP H02207369 · 1990 [cited by applicant]
JP H02267660 · 1990 [cited by applicant]
JP JH02292602 · 1990 [cited by applicant]
JP 3100857 · 1991 [cited by applicant]
JP H03276137 · 1991 [cited by applicant]
JP H04040581 · 1992 [cited by applicant]
JP H04113481 · 1992 [cited by applicant]
JP H04148212 · 1992 [cited by applicant]
JP H04182769 · 1992 [cited by applicant]
JP H04237388 · 1992 [cited by applicant]
JP H04328691 · 1992 [cited by applicant]
JP H04336370 · 1992 [cited by applicant]
JP H04344970 · 1992 [cited by applicant]
JP H0520294A · 1993 [cited by applicant]
JP H05067064 · 1993 [cited by applicant]
JP H0581227 · 1993 [cited by applicant]
JP H05128284 · 1993 [cited by applicant]
JP H05128285 · 1993 [cited by applicant]
JP H05159087 · 1993 [cited by applicant]
JP H05165800 · 1993 [cited by applicant]
JP H05165987 · 1993 [cited by applicant]
JP H05181547 · 1993 [cited by applicant]
JP H05197701 · 1993 [cited by applicant]
JP H05233586A · 1993 [cited by applicant]
JP H05282272 · 1993 [cited by applicant]
JP 5346914 · 1993 [cited by applicant]
JP H05334278 · 1993 [cited by applicant]
JP H0619868 · 1994 [cited by applicant]
JP H06076092 · 1994 [cited by applicant]
JP H06161586 · 1994 [cited by applicant]
JP H06348906A · 1994 [cited by applicant]
JP H07005904 · 1995 [cited by applicant]
JP 7105165 · 1995 [cited by applicant]
JP H07160796 · 1995 [cited by applicant]
JP H07191951 · 1995 [cited by applicant]
JP H07302292 · 1995 [cited by applicant]
JP H08161282 · 1996 [cited by applicant]
JP H09026948 · 1997 [cited by applicant]
JP H1055346A · 1998 [cited by applicant]
JP 2000322400 · 2000 [cited by applicant]
JP 2001034735 · 2001 [cited by applicant]
JP 3177996 · 2001 [cited by applicant]
JP 3183484B2 · 2001 [cited by applicant]
JP 2006154992 · 2006 [cited by applicant]
JP 4264664 · 2009 [cited by applicant]
JP 2009288908 · 2009 [cited by applicant]
JP 2015195011 · 2015 [cited by applicant]
JP 2016153984 · 2016 [cited by applicant]
JP 2016219011 · 2016 [cited by applicant]
KR 2001095960 · 2001 [cited by applicant]
KR 367715B1 · 2003 [cited by applicant]
KR 2005042871 · 2005 [cited by applicant]
KR 919572B1 · 2009 [cited by applicant]
RU 2057363 · 1996 [cited by applicant]
RU 2473126 · 2013 [cited by applicant]
RU 2602973 · 2016 [cited by applicant]
UA 96148C2 · 2011 [cited by applicant]
WO 1991019267A1 · 1991 [cited by applicant]
WO 1993023814A1 · 1993 [cited by applicant]
WO 1996041277A1 · 1996 [cited by applicant]
WO 2005051189A2 · 2005 [cited by applicant]
WO 2007033101A2 · 2007 [cited by applicant]
WO 2012006468A1 · 2012 [cited by applicant]
WO 2015053889A2 · 2015 [cited by applicant]
WO 2015157013A1 · 2015 [cited by applicant]
WO 2015193531A1 · 2015 [cited by applicant]
WO 2016099779A1 · 2016 [cited by applicant]
WO 2016186811A1 · 2016 [cited by applicant]
WO WO2018126270A1 · 2018 [cited by applicant]
Shuchang Zhou et al., “Balanced Quantization: An Effective and Efficient Approach to Quantized Neural Networks”, arXiv, Jun. 2017 (Year: 2017). [cited by applicant]
Lin, “Fixed Point Quantization of Deep Convolutional Networks”, Proceedings of the 33rd International Conference on Machine Learning, New York, NY, JMLR: W&CP vol. 48, arXiv, Jun. 2016 (Year: 2016). [cited by applicant]
Amir et al., “Cognitive Computing Programming Paradigm: A Corelet Language for Composing Networks ofNeurosynaptic Cores”, Aug. 4-9, 2013, The 2013 International Joint Conference on Neural Networks (IJCNN), IEEE (Rear: 2… [cited by applicant]
George et al.,MicroPython 1.14 [Online}, Aug. 2015 [retrived on Mar. 16, 2021], Retrieved from the Internet:,URL: http://docs.micropython.org/en/latest/pyboard/tutorial/assembler.html. (Year: 2015). [cited by applicant]
Pattnaik et al., “Scheduling Techniques for GPU Architectures with Processing-In_Memory Capabilities”, Sep. 11-15, 2016, IEEE, p. 31-44 (Year: 2016). [cited by applicant]
M'zah et al., “Deterministic Microcode Machine Generation”, 2017 IEEE Conference on Dependable and Secure Computing, 2017, pp. 370-376, doi: 10.1109/DESEC.2017.8073858. (Year: 2017). [cited by applicant]
ISR/WO dated Jun. 24, 2018; issued in connection with PCT/IL2018/050392. [cited by applicant]
ISR/WO dated Jul. 10, 2018; issued in connection with PCT/IL2018/050393. [cited by applicant]
ISR/WO dated Jul. 12, 2018; issued in connection with PCT/IL2018/050394. [cited by applicant]
ISR/WO dated Jul. 11, 2018; issued in connection with PCT/IL2018/050395. [cited by applicant]
ISR/WO dated Jul. 15, 2018; issued in connection with PCT/IL2018/050396. [cited by applicant]
Yihua Liao: Neural Networks in Hardware: A Survey (undated). [cited by applicant]
Zhu et al., “Flow-Guided Feature Aggregation for Video Object Detection”, Arxiv Technology Report, arXiv:1703.10025v2 [cs.CV], Aug. 2017. [cited by applicant]
Campolucci et al., “On-Line Learning Algorithms for Locally Recurrent Neural Networks”, IEEE Transactions on Neural Networks, vol. 1 , No. 2, pp. 253-271, Mar. 1999. [cited by applicant]
Kumpati S. Narendra et al., “Identification and Control of Dynamical Systems Using Neural Networks”, IEEE Transactions on Neural Networks, vol. 1, No. 1, pp. 4-27, Mar. 1990. [cited by applicant]
Principe et al., “An Analysis of the Gamma Memory in Dynamic Neural Networks”, IEEE Transactions on Neural Networks, vol. 5, No. 2, pp. 331-337, Mar. 1994. [cited by applicant]
Khan M et al, SpinNaker: Mapping neural networks onto a massively-parallel chip multiprocessor, Neural Networks, 2008, pp. 2849-2856. [cited by applicant]
Carillo et al, Hierarchical network-on-chip and traffic ompression . . . , 2012 IEEE, p. 83-90. [cited by applicant]
V.T. Lee, A. Alaghi, J. P. Hayes, V. Sathe and L. Ceze, “Energy-efficient hybrid stochastic-binary neural networks for near-sensor computing,” [cited by applicant]
Makato Yamada, Sparse Network Lasso for Local High-dimensional Regression, Mar. 23, 2016, p. 1-12 (Year: 2016). [cited by applicant]
Sze, Vivienne et al. “Efficient Processing of Deep Neural Networks: A Tutorial and Survey.” Proceedings of the IEEE 105 (2017): 2295-2329. [cited by applicant]
Wei Wen, Learning Structured Sparsity in Deep Neural Networks, Dec. 5, 2016, NIPS'16: Proceedings of the 30th International Conference on Neural Information Processing Systems, p. 1-8. (Year: 2016). [cited by applicant]
Majid, Mohammad Wadood, Golrokh Mirzaei, and Mohsin M. Jamali. “Evolutionary Neural Network parallelization with multicore systems on chip.” 2012 IEEE International Conference on Electro/Information Technology. IEEE, 20… [cited by applicant]
Cited By (2)
US 12,646,584 US 12,670,432