US Patent Application 17715884. SCALABLE DEEP LEARNING DESIGN FOR MISSING INPUT FEATURES simplified abstract
Contents
SCALABLE DEEP LEARNING DESIGN FOR MISSING INPUT FEATURES
Inventors
Mohamed Fouad Ahmed MARZBAN of San Diego CA (US)
Wooseok NAM of San Diego CA (US)
Taesang YOO of San Diego CA (US)
Mahmoud TAHERZADEH BOROUJENI of San Diego CA (US)
Arumugam CHENDAMARAI KANNAN of San Diego CA (US)
SCALABLE DEEP LEARNING DESIGN FOR MISSING INPUT FEATURES - A simplified explanation of the abstract
- This abstract for appeared for patent application number 17715884 Titled 'SCALABLE DEEP LEARNING DESIGN FOR MISSING INPUT FEATURES'
Simplified Explanation
This abstract describes a method of wireless communication using a neural network. The method involves determining if the input received for each input branch of the neural network is conforming or non-conforming. Each input branch represents a different type of input features. If non-conforming input is detected, the method replaces the activation of certain layers in the neural network. Finally, the method predicts an output parameter based on the conforming input and the replacement activation.
Original Abstract Submitted
A method of wireless communication, by a user equipment (UE), includes determining whether conforming input has been received for each of multiple input branches to a neural network. Each input branch represents a different modality of input features to the neural network. The method also includes replacing an activation with a replacement activation for at least one neural network layer of each of the input branches associated with input determined to be non-conforming input. The method further includes predicting an output parameter based on conforming input received at each input branch and the replacement activation.
