Determining perturbation mask for classification model

A classification model and mask technology, applied in the system field, can solve problems such as real generalization performance problems of classification models

Pending Publication Date: 2020-11-17
ROBERT BOSCH GMBH
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Also in this case, the true generalization performance of the classification model may be questioned

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Determining perturbation mask for classification model
  • Determining perturbation mask for classification model
  • Determining perturbation mask for classification model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0049] figure 1 A system 100 for determining a perturbation mask is shown. A perturbation mask may indicate perturbations for an input instance to a classification model that interfere with classification of the input instance by the classification model. System 100 may include input interface 120 and processor subsystem 140 , which may be in internal communication via data communication 124 .

[0050] Processor subsystem 140 may be configured to access classification model 040 and generative model 060 during operation of system 100 and using input interface 120 . Classification Model 040 may be configured to determine a classification of a certain type of input instance. The generative model 060 may be configured to generate synthetic instances of this type from latent space representations. For example, if figure 1 As shown in , a data interface 120 may provide access 122 to an external data store 020 which may include said models 040 , 060 . Alternatively, models 040 ,...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses determining perturbation mask for a classification model. A system (100) is disclosed for determining, for an input instance to a classification model, a mask indicating perturbations that disturb a classification of the input instance by the classification model. A classification model determines classifications of input instances of a certain type. A generative model generates synthetic instances of the type from latent space representations. Given an input instance, its classification according to the classification model, and a latent space representation that letsthe generative model approximate the input instance, the mask is determined. The mask indicates perturbations to the latent space representation for the input instance and is determined based on a classification score of the classification model for a perturbed input instance. The perturbed instance is determined using the mask by masking the latent space representation with the mask and generating the perturbed input instance from the masked latent space representation using the generative model.

Description

technical field [0001] The present invention relates to a system for determining masks for classification by a classification model, eg for gaining an understanding of whether the classification model has learned input / output relations. The invention also relates to a corresponding computer-implemented method. The invention furthermore relates to a computer-readable medium comprising instructions for performing the method. [0002] Background of the invention [0003] Classification models are increasingly used in contexts such as automated control systems. For example, an autonomous vehicle's control system may use image classification to detect objects in their vicinity, such as traffic signs and obstacles, and use this detection to control the vehicle. However, there are also non-autonomous vehicles today that have driver assistance systems that use image classification, eg for braking if the vehicle is in danger of colliding with an object. Image classification is also...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06N3/02G06N3/08
CPCG06N3/02G06N3/08G06F18/24323G06F18/2415G06F18/24G06F18/214G06F18/2413G06F18/2433
Inventor A.M.蒙诺兹德尔加多
Owner ROBERT BOSCH GMBH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products