Patents by Inventor Pegah Ghahremani

Pegah Ghahremani has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).

  • Self-stabilized deep neural network
    Patent number: 10885438
    Abstract: A neural network is structured with a plurality of levels of nodes. Each level has a level-specific stabilization parameter that adjusts a learning rate, at a corresponding level, during training. The stabilization parameter has a value that varies inversely relative to a change in an objective training function during back-propagation of the error through the level.
    Type: Grant
    Filed: December 28, 2015
    Date of Patent: January 5, 2021
    Assignee: Microsoft Technology Licensing, LLC
    Inventors: James G. Droppo, Pegah Ghahremani, Avner May
  • Linearly augmented neural network
    Patent number: 10558909
    Abstract: A neural network is structured to connect the input values of an input set, at each level, to that level's output using a linear bypass connection. The linear bypass connection passes the input values, to the output, without applying a non-linear function to them.
    Type: Grant
    Filed: December 28, 2015
    Date of Patent: February 11, 2020
    Assignee: Microsoft Technology Licensing, LLC
    Inventors: James G. Droppo, Pegah Ghahremani, Michael L. Seltzer
  • LINEARLY AUGMENTED NEURAL NETWORK
    Publication number: 20170185887
    Abstract: A neural network is structured to connect the input values of an input set, at each level, to that level's output using a linear bypass connection. The linear bypass connection passes the input values, to the output, without applying a non-linear function to them.
    Type: Application
    Filed: December 28, 2015
    Publication date: June 29, 2017
    Inventors: James G. Droppo, Pegah Ghahremani, Michael L. Seltzer
  • SELF-STABILIZED DEEP NEURAL NETWORK
    Publication number: 20170185897
    Abstract: A neural network is structured with a plurality of levels of nodes. Each level has a level-specific stabilization parameter that adjusts a learning rate, at a corresponding level, during training.
    Type: Application
    Filed: December 28, 2015
    Publication date: June 29, 2017
    Inventors: James G. Droppo, Pegah Ghahremani, Avner May