Patents by Inventor Jaakko T. KARRAS

Jaakko T. KARRAS 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).

  • ENHANCED NEURAL NETWORK TRAINING VIA FORCED NORMALIZATION OF LEARNED WEIGHTS
    Publication number: 20250111233
    Abstract: Apparatuses, systems, and techniques to train neural networks. In at least one embodiment, a first normalization of learned parameters of one or more learned layers is performed during a forward pass of a training iteration and a second normalization of the learned parameters is performed during a parameter update phase of the training iteration. In at least one embodiment, the first normalization is performed using first scaling factors and the second normalization is performed using second scaling factors.
    Type: Application
    Filed: September 26, 2024
    Publication date: April 3, 2025
    Inventors: Tero Tapani Karras, Miika Samuli Aittala, Janne Johannes Hellsten, Jaakko T. Lehtinen, Timo Oskari Aila, Samuli Matias Laine
  • ENHANCING NEURAL NETWORK TRAINING VIA BALANCED CONCATENATION
    Publication number: 20250111227
    Abstract: Apparatuses, systems, and techniques to train neural networks and to use neural networks to perform inference. In at least one embodiment, a balanced concatenation layer performs a balanced concatenation operation during a forward pass of a training iteration during the training of a neural network. In at least one embodiment, a balanced concatenation layer performs a balanced concatenation operation during the use of a neural network to perform inference.
    Type: Application
    Filed: September 26, 2024
    Publication date: April 3, 2025
    Inventors: Tero Tapani Karras, Miika Samuli Aittala, Janne Johannes Hellsten, Jaakko T. Lehtinen, Timo Oskari Aila, Samuli Matias Laine
  • ENHANCING NEURAL NETWORK TRAINING AND DEVELOPMENT VIA POST-HOC PARAMETER AVERAGING
    Publication number: 20250111245
    Abstract: Apparatuses, systems, and techniques to compute neural network parameters and to use a neural network to perform inference. In at least one embodiment, neural network parameters are computed, after training, by determining a weighted average of snapshots of averaged parameters that form a basis set of averaged parameter snapshots, each respective snapshot of averaged parameters including a plurality of network parameters averaged by a respective combination of an averaging function and one or more averaging parameters.
    Type: Application
    Filed: September 26, 2024
    Publication date: April 3, 2025
    Inventors: Samuli Matias Laine, Miika Samuli Aittala, Janne Johannes Hellsten, Jaakko T. Lehtinen, Timo Oskari Aila, Tero Tapani Karras
  • Wearable electromyography sensor array using conductive cloth electrodes for human-robot interactions
    Patent number: 10449672
    Abstract: A sleeve worn on an arm allows detection of gestures by an array of sensors. Electromyography, inertial, and magnetic field sensors provide data that is processed to categorize gestures and translate the gestures into commands for robotic systems. Machine learning allows training of gestures to increase accuracy of detection for different users.
    Type: Grant
    Filed: February 28, 2017
    Date of Patent: October 22, 2019
    Assignee: CALIFORNIA INSTITUTE OF TECHNOLOGY
    Inventors: Christopher Assad, Jaakko T. Karras, Michael T. Wolf, Adrian Stoica
  • Puffer: pop-up flat folding explorer robot
    Patent number: 10106214
    Abstract: A repeatably reconfigurable robot, comprising at least two printed circuit board (PCB) rigid sections, at least one PCB flexible section coupled to the at least two PCB rigid sections, at least one wheel, hybrid wheel propeller, wheel and propeller, or hybrid wheel screw propeller rotatably coupled to at least one of the at least two PCB rigid sections and at least one actuator coupled to the at least two PCB rigid sections, wherein the at least one actuator folds and unfolds the repeatably reconfigurable robot.
    Type: Grant
    Filed: September 21, 2016
    Date of Patent: October 23, 2018
    Assignee: CALIFORNIA INSTITUTE OF TECHNOLOGY
    Inventors: Jaakko T. Karras, Christine Fuller, Kalind C. Carpenter, Alessandro Buscicchio, Carolyn E. Parcheta
  • WEARABLE ELECTROMYOGRAPHY SENSOR ARRAY USING CONDUCTIVE CLOTH ELECTRODES FOR HUMAN-ROBOT INTERACTIONS
    Publication number: 20170259428
    Abstract: A sleeve worn on an arm allows detection of gestures by an array of sensors. Electromyography, inertial, and magnetic field sensors provide data that is processed to categorize gestures and translate the gestures into commands for robotic systems. Machine learning allows training of gestures to increase accuracy of detection for different users.
    Type: Application
    Filed: February 28, 2017
    Publication date: September 14, 2017
    Inventors: Christopher ASSAD, Jaakko T. KARRAS, Michael T. WOLF, Adrian STOICA
  • PUFFER: POP-UP FLAT FOLDING EXPLORER ROBOT
    Publication number: 20170088205
    Abstract: A repeatably reconfigurable robot, comprising at least two printed circuit board (PCB) rigid sections, at least one PCB flexible section coupled to the at least two PCB rigid sections, at least one wheel, hybrid wheel propeller, wheel and propeller, or hybrid wheel screw propeller rotatably coupled to at least one of the at least two PCB rigid sections and at least one actuator coupled to the at least two PCB rigid sections, wherein the at least one actuator folds and unfolds the repeatably reconfigurable robot.
    Type: Application
    Filed: September 21, 2016
    Publication date: March 30, 2017
    Inventors: Jaakko T. KARRAS, Christine FULLER, Kalind C. CARPENTER, Alessandro BUSCICCHIO, Carolyn E. PARCHETA