Robust neural network learning system

Patent metadata

Field	Value
Publication number	US-12243295-B2
Application number	US-202217709553-A
Country	US
Kind code	B2
Filing date	Mar 31, 2022
Priority date	Mar 31, 2022
Publication date	Mar 4, 2025
Grant date	Mar 4, 2025

Abstract

A system comprises a computer including a processor and a memory. The memory includes instructions such that the processor is programmed to: receive intermediate concept constraints at a neural network and train the neural network with training data, training labels, and the at least one of the data constraint, the feature constraint, or the intermediate concept constraint.

First claim

What is claimed is:

1. A system comprising a computer including a processor and a memory, the memory including instructions such that the processor is programmed to:
receive a data constraint, a feature constraint, and an intermediate concept constraint at a neural network, wherein the data constraint includes a perturbed image of an object, the feature constraint includes a physical property of a vehicle sensor, and the intermediate concept constraint is based on human knowledge that provides additional context about a training data, the additional context comprising additional definitions and relationships pertaining to an object depicted within the training data; and
train the neural network with the training data, training labels, and the data constraint, the feature constraint, and the intermediate concept constraint.

2. The system of claim 1 , wherein the intermediate concept constraint further includes at least one concept parameter that defines an individual component and relational information pertaining to an object of interest.

3. The system of claim 1 , wherein the feature constraint further comprises style parameters corresponding to the input data.

4. The system of claim 3 , wherein the feature constraint further comprises intrinsic image parameters corresponding to sensor data and extrinsic image parameters corresponding to the sensor data.

5. The system of claim 1 , wherein the processor is further programmed to receive the training data and the training labels.

6. The system of claim 1 , wherein the training data comprises images depicting the object within a field-of-view of the vehicle sensor.

7. The system of claim 1 , wherein the neural network comprises a deep neural network.

8. The system of claim 7 , wherein the deep neural network comprises at least one of a convolutional neural network or a generative adversarial neural network.

9. A method comprising:
receiving a data constraint, a feature constraint, and an intermediate concept constraint at a neural network, wherein the data constraint includes a perturbed image of an object, the feature constraint includes a physical property of a vehicle sensor, and the intermediate concept constraint is based on human knowledge that provides additional context about a training data, the additional context comprising additional definitions and relationships pertaining to an object depicted within the training data; and
training the neural network with the training data, training labels, and the data constraint, the feature constraint, and the intermediate concept constraint.

10. The method of claim 9 , wherein the intermediate concept constraint further includes at least one concept parameter that defines an individual component and relational information pertaining to an object of interest.

11. The method of claim 9 , wherein the feature constraint further comprises style parameters corresponding to the input data.

12. The method of claim 11 , wherein the feature constraint further comprises intrinsic image parameters corresponding to sensor data and extrinsic image parameters corresponding to the sensor data.

13. The method of claim 9 , further comprising receiving the training data and the training labels.

14. The method of claim 9 , wherein the training data comprises images depicting the object within a field-of-view of the vehicle sensor.

15. The method of claim 9 , wherein the neural network comprises a deep neural network.

16. The method of claim 15 , wherein the deep neural network comprises at least one of a convolutional neural network or a generative adversarial neural network.

Assignees

• Gm Global Tech Operations Llc

Inventors

• Tong Wei
• Bond Jacob Alan
• Gupta Siddhartha

Classifications

• G06N3/08 Primary

  Learning methods · CPC title

• G06V10/82 Primary

  using neural networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/094

  Adversarial learning · CPC title

• G06N3/063

  using electronic means · CPC title