Processing

Please wait...

Settings

Settings

Goto Application

1. WO2020242689 - EXECUTION OF DEEP-LEARNING MODEL

Publication Number WO/2020/242689
Publication Date 03.12.2020
International Application No. PCT/US2020/030018
International Filing Date 27.04.2020
IPC
G06N 3/04 2006.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architecture, e.g. interconnection topology
G06F 21/50 2013.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
FELECTRIC DIGITAL DATA PROCESSING
21Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
50Monitoring users, programs or devices to maintain the integrity of platforms, e.g. of processors, firmware or operating systems
CPC
G06F 21/53
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
FELECTRIC DIGITAL DATA PROCESSING
21Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
50Monitoring users, programs or devices to maintain the integrity of platforms, e.g. of processors, firmware or operating systems
52during program execution, e.g. stack integrity ; ; Preventing unwanted data erasure; Buffer overflow
53by executing in a restricted environment, e.g. sandbox or secure virtual machine
G06N 3/0454
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architectures, e.g. interconnection topology
0454using a combination of multiple neural nets
G06N 3/0481
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architectures, e.g. interconnection topology
0481Non-linear activation functions, e.g. sigmoids, thresholds
G06N 3/063
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
06Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
063using electronic means
Applicants
  • MICROSOFT TECHNOLOGY LICENSING, LLC [US]/[US]
Inventors
  • LIU, Yunxin
  • ZHANG, Lintao
Agents
  • SWAIN, Cassandra T.
  • BARKER, Doug
  • CHATTERJEE, Aaron C.
  • CHEN, Wei-Chen Nicholas
  • CHOI, Daniel
  • CHURNA, Timothy
  • DINH, Phong
  • EVANS, Patrick
  • GABRYJELSKI, Henry
  • GUPTA, Anand
  • HINOJOSA-SMITH, Brianna L.
  • HWANG, William C.
  • JARDINE, John S.
  • LEE, Sunah
  • LEMMON, Marcus
  • MARQUIS, Thomas
  • MEYERS, Jessica
  • ROPER, Brandon
  • SPELLMAN, Steven
  • SULLIVAN, Kevin
  • WALKER, Matt
  • WIGHT, Stephen A.
  • WISDOM, Gregg
  • WONG, Ellen
  • WONG, Thomas S.
  • ZHANG, Hannah
  • TRAN, Kimberly
Priority Data
201910475938.131.05.2019CN
Publication Language English (EN)
Filing Language English (EN)
Designated States
Title
(EN) EXECUTION OF DEEP-LEARNING MODEL
(FR) EXÉCUTION D'UN MODÈLE D'APPRENTISSAGE PROFOND
Abstract
(EN)
In accordance with implementations of the subject matter described herein, there is provided a solution for execution of a deep learning model. In the solution, partitioned convolutions are executed based on an input and a set of parameter values of the convolutional layer sequentially in a trusted execution environment (TEE) of a computing device. The execution of a given one of partitioned convolutions comprises: storing, into a protected memory area in the TEE, an input portion of the input to be processed by a subset of parameter values for the given partitioned convolution; determining a result of the given partitioned convolution through a single matrix multiplication operation; and removing the input portion. By combining results of the partitioned convolutions, a result of the convolution is determined. Therefore, the solution can accelerate the model execution speed and improve the storage efficiency in a highly safe TEE with limited memory resources.
(FR)
Des modes de réalisation de la présente invention concernent une solution pour l'exécution d'un modèle d'apprentissage profond. Dans la solution, des convolutions partitionnées sont exécutées sur la base d'une entrée et d'un ensemble de valeurs de paramètre de la couche de convolution séquentiellement dans un environnement d'exécution de confiance (TEE) d'un dispositif informatique. L'exécution de l'une donnée des convolutions partitionnées consiste à : stocker, dans une zone de mémoire protégée dans le TEE, une portion d'entrée de l'entrée devant être traitée par un sous-ensemble de valeurs de paramètre pour la convolution partitionnée donnée ; déterminer un résultat de la convolution partitionnée donnée via une opération de multiplication à matrice unique ; et supprimer la portion d'entrée. La combinaison des résultats des convolutions partitionnées permet de déterminer un résultat de la convolution. Par conséquent, la solution peut accélérer la vitesse d'exécution du modèle et améliorer l'efficacité de stockage dans un TEE très sûr avec des ressources de mémoire limitées.
Also published as
Latest bibliographic data on file with the International Bureau