Traitement en cours

Veuillez attendre...

Paramétrages

Paramétrages

Aller à Demande

1. WO2020109774 - VÉRIFICATION DE SYSTÈMES DE PERCEPTION

Note: Texte fondé sur des processus automatiques de reconnaissance optique de caractères. Seule la version PDF a une valeur juridique

[ EN ]

Claims

1. A computer-implemented method for verifying the robustness of a neural network classifier with respect to one or more parameterised transformations applied to an input, the classifier comprising one or more convolutional layers, the method comprising:

encoding each layer of the classifier as one or more algebraic classifier constraints; encoding each transformation as one or more algebraic transformation

constraints;

encoding a change in an output classifier label from the classifier as an algebraic output constraint;

determining whether a solution exists which satisfies the classifier constraints, transformation constraints and output constraints, and determining the classifier as robust to the local transformations if no such solution exists.

2. A method according to claim l, further comprising, where a solution exists which satisfies the classifier constraints, transformation constraints and output constraints: identifying parameters of the one or more transformations associated with the solution.

3 A method according to claim 2, comprising:

generating additional training data in dependence on the identified parameters; training the classifier using the training data.

4. A method according to claim 3, wherein generating the additional training data comprises applying the one or more transformations to existing training data using the identified parameters.

5. A method according to any one of the preceding claims, wherein one or more of the classifier, transformation and output constraints are linear constraints.

6. A method according to any one of the preceding claims, wherein one or more of the classifier, transformation and output constraints are non-linear constraints.

7. A method according to any one of the preceding claims, wherein at least one of the one or more local transformations is a geometric transformation.

8. A method according to any one of the preceding claims, wherein at least one of the one or more local transformations is a photometric transformation.

9. A method according to any one of the preceding claims, wherein at least one of the one or more local transformations is an affine transformation.

10. A method according to any one of the preceding claims, wherein the classifier further comprises one or more fully connected layers.

11. A method according to any one of the preceding claims, wherein encoding each layer of the classifier as one or more algebraic classifier constraints comprises deriving a mixed-integer linear programming expression for each layer.

12. A method according to any one of the preceding claims, wherein encoding each transformation as one or more algebraic classifier constraints comprises deriving a mixed-integer linear programming expression for each layer.

13. A method according to any one of the preceding claims, wherein encoding each transformation as one or more algebraic classifier constraints comprises deriving a mixed-integer non-linear programming expression for each layer.

14. A method according to any one of the preceding claims, wherein one or more of the classifier layers comprises a rectified linear unit activating function.

15. A method according to any one of the preceding claims, wherein the classifier is an image classifier.

16. A method according to any one of claims 1 to 15, wherein the classifier is an audio classifier.

17. A computer program product comprising computer executable instructions which, when executed by one or more processors, cause the one or more processors to carry out the method of any one of the preceding claims.

18. A perception system comprising one or more processors configured to carry out the method of any one of claims l to 17.