FR3090953B1 - Method for checking the robustness of a neural network - Google Patents

Method for checking the robustness of a neural network Download PDF

Info

Publication number
FR3090953B1
FR3090953B1 FR1873742A FR1873742A FR3090953B1 FR 3090953 B1 FR3090953 B1 FR 3090953B1 FR 1873742 A FR1873742 A FR 1873742A FR 1873742 A FR1873742 A FR 1873742A FR 3090953 B1 FR3090953 B1 FR 3090953B1
Authority
FR
France
Prior art keywords
neural network
critical
input signal
robustness
checking
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
FR1873742A
Other languages
French (fr)
Other versions
FR3090953A1 (en
Inventor
Sergey Abrashov
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
PSA Automobiles SA
Original Assignee
PSA Automobiles SA
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by PSA Automobiles SA filed Critical PSA Automobiles SA
Priority to FR1873742A priority Critical patent/FR3090953B1/en
Publication of FR3090953A1 publication Critical patent/FR3090953A1/en
Application granted granted Critical
Publication of FR3090953B1 publication Critical patent/FR3090953B1/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Traffic Control Systems (AREA)
  • Testing And Monitoring For Control Systems (AREA)
  • Image Analysis (AREA)

Abstract

L’invention concerne un procédé de validation d’un système de réseau neuronal pour un système, notamment pour un véhicule, le réseau neuronal étant configuré pour prendre une décision de réalisation d’une action et recevant en entrée au moins un signal issu d’au moins un capteur du système, ledit procédé comprenant les étapes suivantes : la détermination (E1) d’au moins une sortie critique du réseau neuronal, engendrant la non-réalisation de l’action, la détermination (E2) d’au moins un signal d’entrée, dit critique, correspondant à ladite au moins une situation critique, la vérification (E3) que ledit au moins un signal d’entrée critique correspond à un signal d’entrée déterminé comme possible, autrement dit susceptible d’être généré par ledit au moins un capteur correspondant en situation réelle d’utilisation du système. Figure de l’abrégé : Figure 2The invention relates to a method for validating a neural network system for a system, in particular for a vehicle, the neural network being configured to take a decision to carry out an action and receiving as input at least one signal originating from at least one sensor of the system, said method comprising the following steps: the determination (E1) of at least one critical output of the neural network, causing the non-performance of the action, the determination (E2) of at least one input signal, called critical, corresponding to said at least one critical situation, the verification (E3) that said at least one critical input signal corresponds to an input signal determined as possible, in other words capable of being generated by said at least one corresponding sensor in a real situation of use of the system. Abstract figure: Figure 2

FR1873742A 2018-12-21 2018-12-21 Method for checking the robustness of a neural network Expired - Fee Related FR3090953B1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
FR1873742A FR3090953B1 (en) 2018-12-21 2018-12-21 Method for checking the robustness of a neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
FR1873742A FR3090953B1 (en) 2018-12-21 2018-12-21 Method for checking the robustness of a neural network

Publications (2)

Publication Number Publication Date
FR3090953A1 FR3090953A1 (en) 2020-06-26
FR3090953B1 true FR3090953B1 (en) 2020-12-04

Family

ID=66542431

Family Applications (1)

Application Number Title Priority Date Filing Date
FR1873742A Expired - Fee Related FR3090953B1 (en) 2018-12-21 2018-12-21 Method for checking the robustness of a neural network

Country Status (1)

Country Link
FR (1) FR3090953B1 (en)

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2941352C (en) 2014-03-06 2022-09-20 Progress, Inc. Neural network and method of neural network training
US10133275B1 (en) * 2017-03-01 2018-11-20 Zoox, Inc. Trajectory generation using temporal logic and tree search

Also Published As

Publication number Publication date
FR3090953A1 (en) 2020-06-26

Similar Documents

Publication Publication Date Title
US10545228B2 (en) Object identification device
RU2008101894A (en) TRAINING TRAINING SYSTEMS USING RANDOM FUNCTIONS OF COST
WO2015146082A1 (en) Leak-detecting device, leak detection method, and program-containing recording medium
US20130291610A1 (en) Apparatus and method for measuring the bending angle of a sheet
DK2053539T3 (en) Classifying objects as well as detecting their position and orientation in space
FR3036838B1 (en) ELECTRONIC WIND MUSICAL INSTRUMENT
CN111475410B (en) Test case generation method and device
EA201201369A1 (en) METHOD FOR PROCESSING BANKNOTES (OPTIONS)
CN110738108A (en) Target object detection method, target object detection device, storage medium and electronic equipment
CN104915293A (en) Software testing method and system
FR3090953B1 (en) Method for checking the robustness of a neural network
FR3059802B1 (en) METHOD FOR GENERATING AN ELECTRONIC SIGNATURE OF A DOCUMENT ASSOCIATED WITH A CONDENSATE
FR3069365B1 (en) METHOD AND SYSTEM FOR DETECTION OF A FREE ZONE IN A CAR PARK
KR20130129991A (en) Method for calibrating a modulation point of a pneumatic brake booster which is calculated on the basis of a pressure sensor signal
FR3091589B1 (en) Method of checking a damage tolerance property of a part made of an aluminum alloy
FR3079026B1 (en) METHOD OF CALIBRATION OF A GYROMETER EQUIPPING A VEHICLE
FR3069366B1 (en) SPECIFIC ENVIRONMENTAL INPUT PROTECTION
Kettl et al. The static analyzer Infer in SV-COMP (competition contribution)
US8689327B2 (en) Method for characterization of a computer program part
US11656166B2 (en) Method and apparatus for detecting particles
FR3025601B1 (en) VEHICLE CALCULATOR AND SENSOR TYPE DETECTION METHOD OF CRANKSHAFT SENSOR
FR2985074A1 (en) METHOD AND DEVICE FOR DETERMINING A LATERAL TRACK OF AN AIRCRAFT AND ASSOCIATED FLIGHT MANAGEMENT SYSTEM
Lee et al. Density measurement algorithm for freeway segment using two point detectors
FR3073644B1 (en) IMAGE PROCESSING PROCESS IMPLEMENTED BY A TERMINAL FORMING A "WHITE BOX" ENVIRONMENT
FR3094798B1 (en) method of determining a speed of a motor vehicle

Legal Events

Date Code Title Description
PLFP Fee payment

Year of fee payment: 2

PLSC Search report ready

Effective date: 20200626

ST Notification of lapse

Effective date: 20210806