US Patent Application 18309335. TRAINING NEURAL NETWORKS WITH A LESSER REQUIREMENT FOR LABELLED TRAINING DATA simplified abstract

From WikiPatents
Jump to navigation Jump to search

TRAINING NEURAL NETWORKS WITH A LESSER REQUIREMENT FOR LABELLED TRAINING DATA

Organization Name

Robert Bosch GmbH


Inventor(s)

Piyapat Saranrittichai of Nuernberg (DE)

Andres Mauricio Munoz Delgado of Schoenaich (DE)

Chaithanya Kumar Mummadi of Pittsburgh PA (US)

Claudia Blaiotta of Stuttgart (DE)

Volker Fischer of Renningen (DE)

TRAINING NEURAL NETWORKS WITH A LESSER REQUIREMENT FOR LABELLED TRAINING DATA - A simplified explanation of the abstract

This abstract first appeared for US patent application 18309335 titled 'TRAINING NEURAL NETWORKS WITH A LESSER REQUIREMENT FOR LABELLED TRAINING DATA

Simplified Explanation

The patent application describes a method for training a neural network to determine a task output based on a given task.

  • The method involves providing training records of measurement data, which can be either unlabeled or labeled.
  • The encoder network is trained to map these training records to representations that fulfill a self-consistency condition or correspond to ground truth.
  • Task training records that are labeled with ground truth are also provided.
  • The association network and task head networks are trained to map the representations obtained from the encoder network to task outputs that correspond to the ground truth.
  • The training process is guided by a task loss function, which measures the accuracy of the obtained task outputs compared to the labeled ground truth.


Original Abstract Submitted

A method for training a neural network for determining a task output with respect to a given task. The method includes: providing unlabeled and/or labelled encoder training records of measurement data; training the encoder network to map encoder training records to representations towards the goal that these representations, and/or or one or more work products derived from the representations, fulfil a self-consistency condition or correspond to ground truth; providing task training records that are labelled with ground truth; and training the association network and the task head networks towards the goal that, when a task training record is mapped to a representation using the encoder network, and the representation is mapped to a task output by the combination of the association network and the task head networks, the so-obtained task output corresponds to the ground truth with which the training record is labelled, as measured by a task loss function.

Retrieved from "http://wikipatents.org/index.php?title=US_Patent_Application_18309335._TRAINING_NEURAL_NETWORKS_WITH_A_LESSER_REQUIREMENT_FOR_LABELLED_TRAINING_DATA_simplified_abstract&oldid=7282"