Title:
REFINING MACHINE LEARNING MODELS TO MITIGATE ADVERSARIAL ATTACKS IN AUTONOMOUS SYSTEMS AND APPLICATIONS
Document Type and Number:
WIPO Patent Application WO/2024/098374
Kind Code:
A1
Abstract:
In various examples, a technique for processing sensor data includes generating, using a machine learning model and based on a first sensor data instance, a first set of confidences for a set of output types and a first adversarial confidence that represents a likelihood that the first sensor data instance is adversarial. The technique also includes determining that the first sensor data instance is adversarial based on the first adversarial confidence. The technique further includes transmitting a first indication that the first sensor data instance is adversarial to one or more downstream components such that the one or more downstream components perform one or more operations based at least on the indication.
More Like This:
WO/2024/115204 | CHARGED PARTICLE ASSESSMENT METHOD AND SYSTEM |
JP7045122 | Drone, drone control method, and drone control program |
JP2023018472 | INFORMATION PROCESSING METHOD AND PROGRAM |
Inventors:
YU CHONG (US)
Application Number:
PCT/CN2022/131339
Publication Date:
May 16, 2024
Filing Date:
November 11, 2022
Export Citation:
Assignee:
NVIDIA CORP (US)
YU CHONG (US)
YU CHONG (US)
International Classes:
G06T7/00; G06N20/00; G06T5/00
Foreign References:
US20220261642A1 | 2022-08-18 | |||
US20210286923A1 | 2021-09-16 | |||
US20210406560A1 | 2021-12-30 | |||
US20220092349A1 | 2022-03-24 | |||
US20220215030A1 | 2022-07-07 |
Attorney, Agent or Firm:
P. C. & ASSOCIATES (CN)
Download PDF:
Previous Patent: TECHNIQUES FOR COMPRESSING NEURAL NETWORKS
Next Patent: TECHNIQUES FOR PRUNING NEURAL NETWORKS
Next Patent: TECHNIQUES FOR PRUNING NEURAL NETWORKS