专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明公开

EP4455948A1 COMPUTER-IMPLEMENTIERTES VERFAHREN UND SYSTEM ZUM BETRIEB EINES TECHNISCHEN GERÄTS 审中-公开
公开(公告)号：EP4455948A1
公开(公告)日：2024-10-30
申请号：EP23169732.7
申请日：2023-04-25
申请人： Siemens Aktiengesellschaft
发明人： REZAPOUR LAKANI, Safoura , HIESSL, Thomas , EDER, Jana , SCHALL, Daniel
IPC分类号： G06N3/098 , G06N3/082 , G06N3/09
CPC分类号： G06N3/098 , G06N3/09 , G06N3/082
摘要： Computer-implementiertes Verfahren zum Betreiben eines technischen Geräts (TD) mit einem Modell auf Basis künstlicher Intelligenz durch einen Klienten (C1-C3) eines Klienten-Server-Systems, wobei jeder Klient (C1-C3) ein Klienten-Modell (CM1-CM3) zum Betrieb eines verbundenen technischen Geräts (TD1) speichert, und folgende Schritte ausgeführt werden:
a) Bereitstellen des globalen Modells (GM) auf Basis föderierten Lernens an einen Klienten (C1),
b) Prüfen durch den Klienten (C1-C3), ob im Speicher (MEM) ein sensitiver Modell-Parameter gespeichert ist, welcher nicht vom globalen Modell (GM) umfasst ist, und falls ja, Aggregieren des bereitgestellten globalen Modells (GM) mit dem sensitiven Modell-Parameter, und Aktualisieren des Klienten-Modells (CM1) mit dem bereitgestellten globalen Modell (GM),
c) Bereitstellen eines Referenz-Datensatzes,
d) Berechnen einer ersten Genauigkeit des Klienten-Modells (CM1) mithilfe des Referenz-Datensatzes,
e) Bestimmen der Gradienten der Modell-Parameter des Klienten-Modells (CM1) und Bestimmen eines ausgewählten Gradienten der Modell-Parameter des Klienten-Modells (CM1),
f) Entfernen des Modell-Parameters, welcher zu dem ausgewählten Gradienten zugehörig ist, aus dem Klienten-Modell (CM1),
g) Berechnen einer zweiten Genauigkeit des Klienten-Modells (CM1) mithilfe des Referenz-Datensatzes,
h) Prüfen, ob sich die zweite Genauigkeit unter der ersten Genauigkeit liegt, falls ja, Festlegen des zumindest einen Modell-Parameters, welcher zu dem zumindest einen ausgewählten Gradienten zugehörig ist, als sensitiver Parameter, und Bereitstellen des Klienten-Modells (CM1) an den Server (S)
i) Aktualisieren des globalen Modells,
j) Betreiben des technischen Geräts (TD1) mit dem Klienten-Modell (CM1) und dem sensitiven Parameter.

2. 发明公开

EP4451177A1 GROUP PERSONALIZED FEDERATED LEARNING 审中-公开
公开(公告)号：EP4451177A1
公开(公告)日：2024-10-23
申请号：EP24170582.1
申请日：2024-04-16
申请人： Meta Platforms Technologies, LLC
发明人： Zhe, Liu , Yue, Hui , Fuchun, Peng
IPC分类号： G06N3/09 , G06N3/098
CPC分类号： G06N3/098 , G06N3/09
摘要： A system and method facilitating group personalized federated learning are provided. The system may provide a global machine learning model, generated based on federated learning, to communication devices. The system may also receive model parameters from the communication devices based in part on the communication devices determining local training data generated by the communication devices implementing the global machine learning model. The system may also determine, based on the model parameters from the communication devices, groups of the communication devices and may assign a group specific machine learning model(s), among a plurality of group specific machine learning models, to the groups. The plurality of group specific machine learning models may be associated with the global machine learning model. The system may also provide respective group specific machine learning models, among the plurality of group specific machine learning models, to subsets of communication devices of the groups of the communication devices.

3. 发明公开

EP4446920A1 A METHOD AND SYSTEM FOR IDENTIFICATION OF MALICIOUS CLIENTS WITH SECURE AGGREGATION OF DATA FROM A PLURALITY OF CLIENTS 审中-公开
公开(公告)号：EP4446920A1
公开(公告)日：2024-10-16
申请号：EP23167522.4
申请日：2023-04-12
申请人： Technische Universität München
发明人： Xhemrishi, Marvin , Wachter-Zeh, Antonia , Graell Amat, Alexandre , Östman, Johan
IPC分类号： G06F21/60 , G06F21/62
CPC分类号： G06F21/6245 , G06F21/60 , G06N3/098
摘要： A computer-implemented method for secure aggregation of data from a plurality of clients. The method comprises requesting data from the plurality of clients , and receiving a plurality of masked data portions from the plurality of clients , wherein the plurality of clients are grouped into a plurality of partially overlapping client groups , and each masked data portion is associated with one of the client groups. The method further comprises aggregating the masked data portions of each client group to obtain a plurality of group data aggregates, wherein each group data aggregate is associated with a respective client group , and estimating, based on a quality metric for each of the plurality of group data aggregates, a subset of malicious clients of the plurality of clients. The method further comprises determining a benign data aggregate based on a remaining plurality of benign clients, wherein the benign clients do not form part of the malicious clients.

4. 发明公开

EP4439397A1 SYSTEM AND METHOD FOR CREATING AND EXECUTING SECURED NEURAL NETWORKS 审中-公开
公开(公告)号：EP4439397A1
公开(公告)日：2024-10-02
申请号：EP23166038.2
申请日：2023-03-31
申请人： Irdeto B.V.
发明人： EISEN, Philip Allen , HICKIE, Thomas William , LIN, Tingting
IPC分类号： G06N3/098 , G06N3/0455
CPC分类号： G06N3/098 , G06N3/0455 , G06F21/14 , G06F21/6245
摘要： Disclosed implementations include a method for creating a secured neural network model. A programmatically generated "transcoding" layer can be added to the input and output of an existing neural network model. The transcoding, can be produced using a deterministic algorithm and can leverage known data transforms to protect input data by obfuscating and transforming the input and output data of the model.

5. 发明公开

EP4437661A1 SERVER AND AGENT FOR REPORTING OF COMPUTATIONAL RESULTS DURING AN ITERATIVE LEARNING PROCESS 审中-实审
公开(公告)号：EP4437661A1
公开(公告)日：2024-10-02
申请号：EP21820542.5
申请日：2021-11-26
申请人： Telefonaktiebolaget LM Ericsson (publ)
发明人： MOOSAVI, Reza , RYDÉN, Henrik , LARSSON, Erik, G.
IPC分类号： H04B7/0456 , H04B7/0404 , G06N3/04 , G06N3/08 , G06N20/00
CPC分类号： H04B7/046 , H04B7/0404 , G06N3/098 , G06N3/09 , G06N3/0455

6. 发明公开

EP4433965A1 SERVER AND AGENT FOR REPORTING OF COMPUTATIONAL RESULTS DURING AN ITERATIVE LEARNING PROCESS 审中-实审
公开(公告)号：EP4433965A1
公开(公告)日：2024-09-25
申请号：EP21815149.6
申请日：2021-11-16
申请人： Telefonaktiebolaget LM Ericsson (publ)
发明人： MOOSAVI, Reza , RYDÉN, Henrik , LARSSON, Erik, G.
IPC分类号： G06N20/20 , H04B1/707 , H04L41/16 , H04W52/14 , H04W52/22 , H04W52/24 , H04W52/38
CPC分类号： G06N20/20 , H04L41/16 , H04L5/0055 , H04L41/042 , H04L1/0015 , G06N3/098 , G06N3/0455

7. 发明公开

EP4433950A1 SYSTEM(S) AND METHOD(S) TO REDUCE A TRANSFERABLE SIZE OF LANGUAGE MODEL(S) TO ENABLE DECENTRALIZED LEARNING THEREOF 审中-实审
公开(公告)号：EP4433950A1
公开(公告)日：2024-09-25
申请号：EP23750809.8
申请日：2023-07-06
申请人： GOOGLE LLC
发明人： CHEN, Mingqing , MCCONNAUGHEY, Lara , ÖZGÜN, Kaan Ege , MATHEWS, Rajiv , BEAUFAYS, Françoise
IPC分类号： G06N3/045
CPC分类号： G06N3/098 , G06N3/082 , G06N3/045 , G06N3/09 , G06N3/0895 , G06F40/20 , G06F40/30 , G06N3/044 , G06N3/096

8. 发明公开

EP4433944A1 CIRCUITRY FOR A PLURALITY OF EDGE DEVICES AND METHOD 审中-实审
公开(公告)号：EP4433944A1
公开(公告)日：2024-09-25
申请号：EP22805895.4
申请日：2022-10-26
申请人： Sony Semiconductor Solutions Corporation , Sony Europe B.V.
发明人： MARKHASIN, Lev
IPC分类号： G06N3/006 , G06N3/045 , G06N3/098
CPC分类号： G06N3/098 , G06N3/006 , G06N3/045

9. 发明授权

EP3673419B8 POPULATION BASED TRAINING OF NEURAL NETWORKS 有权转让
公开(公告)号：EP3673419B8
公开(公告)日：2024-09-18
申请号：EP18807623.6
申请日：2018-11-22
IPC分类号： G06N3/098 , G06N3/0985 , G06N3/086 , G06N3/045
CPC分类号： G06N3/086 , G06N3/045 , G06N3/098 , G06N3/0985

10. 发明授权

EP3987456B1 CROSS REPLICA REDUCTION ON NETWORKS HAVING DEGRADED NODES 有权
公开(公告)号：EP3987456B1
公开(公告)日：2024-09-11
申请号：EP20757744.6
申请日：2020-07-30
IPC分类号： G06N3/098 , G06N3/084
CPC分类号： G06N3/084 , G06N3/098

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式