Patent · US Active

Trustworthy predictions using deep neural networks based on adversarial calibration

US11455531B2 · kind B2 · utility

0Cited by

2References

10Claims

0Family size

Assignee

Siemens Aktiengesellschaft · DE

Inventors

Florian Büttner · Munich, DE
Christian Tomani · Munich, DE

Key dates

Filing date	Oct 15, 2019
Grant date	Sep 27, 2022
Priority date	—
Expiry date	Mar 25, 2041

Classification

Technology area (CPC G)Physics
CPC primaryG06V10/82
WIPO fieldComputer technology
WIPO sectorElectrical engineering

Abstract

The disclosed relates to a computer-implemented method of training a Neural Network as well as a corresponding computer program, computer-readable medium and data processing system. In addition to a categorical cross-entropy loss LCCE weights of the NN are updated based on predictive entropy loss LS and an adversarial calibration loss Ladv.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.