利索-全球专利检索

  1. Login
  2. Sign Up

Search Results

220 records (took 0.026 seconds)

    MACHINE LEARNING FIRMWARE OPTIMIZATION
    2.
    发明申请
    MACHINE LEARNING FIRMWARE OPTIMIZATION 有权

    公开(公告)号：US20250045042A1

    公开(公告)日：2025-02-06

    申请号：US18767789

    申请日：2024-07-09

    Applicant: Micron Technology, Inc.

    Inventor： Saideep Tiku ,  Poorna Kale

    IPC: G06F8/658 ,  G06F8/71 ,  G06N3/086 ,  G06N3/092

    Abstract: Machine learning based firmware optimization can include iteratively producing different versions of firmware for operating a physical memory device within a respective defined acceptable range of values for different operational parameters. Iteratively producing different versions of firmware can include deploying an initial version of firmware on a digital twin of the physical memory device, determining an initial value of a performance parameter based on operation of the digital twin according to the initial version of firmware, producing a modified version of firmware, deploying the modified version of firmware on the digital twin, and determining a next value of the performance parameter based on operation of the digital twin according to the modified version of firmware. One of the different versions of firmware that achieves a target value for the performance parameter can be provided for deployment on the physical memory device.

    System and method for efficient evolution of deep convolutional neural networks using filter-wise recombination and propagated mutations
    5.
    发明授权
    System and method for efficient evolution of deep convolutional neural networks using filter-wise recombination and propagated mutations 有权

    公开(公告)号：US12147903B2

    公开(公告)日：2024-11-19

    申请号：US18357554

    申请日：2023-07-24

    Applicant: Nano Dimension Technologies, Ltd.

    Inventor： Eli David

    IPC: G06N3/086 ,  G06N3/045

    Abstract: An efficient technique of machine learning is provided for training a plurality of convolutional neural networks (CNNs) with increased speed and accuracy using a genetic evolutionary model. A plurality of artificial chromosomes may be stored representing weights of artificial neuron connections of the plurality of respective CNNs. A plurality of pairs of the chromosomes may be recombined to generate, for each pair, a new chromosome (with a different set of weights than in either chromosome of the pair) by selecting entire filters as inseparable groups of a plurality of weights from each of the pair of chromosomes (e.g., “filter-by-filter” recombination). A plurality of weights of each of the new or original plurality of chromosomes may be mutated by propagating recursive error corrections incrementally throughout the CNN. A small random sampling of weights may optionally be further mutated to zero, random values, or a sum of current and random values.

    METHOD AND SYSTEM FOR OPTIMIZING PERFORMANCE OF GENETIC ALGORITHM IN SOLVING SCHEDULING PROBLEMS
    7.
    发明申请
    METHOD AND SYSTEM FOR OPTIMIZING PERFORMANCE OF GENETIC ALGORITHM IN SOLVING SCHEDULING PROBLEMS 有权

    公开(公告)号：US20240370730A1

    公开(公告)日：2024-11-07

    申请号：US18766733

    申请日：2024-07-09

    Applicant: Quantiphi, Inc.

    Inventor： Dagnachew Birru ,  Achint Chaudhary ,  Anirudh Deodhar

    IPC: G06N3/086 ,  G06Q10/0631

    Abstract: A method and system for optimizing performance of Genetic Algorithm (GA) in solving scheduling problem is disclosed. The method includes receiving input constraints associated with supply and demand sides, for scheduling problem. The method include initializing set of schedules using initializer that sets initial set of solutions for GA to start optimization. The method may include generating parent population for GA. The method may include creating child population via evolution using current probabilistic parameters including crossover and mutation operators. The method may include utilizing a Multi-Level Hierarchical Grouping (MLHG) to de-duplicate child population. The method includes determining a new population from a total population including the parent population and the child population, using custom multi-objective sorting technique. The method may further include updating probabilistic parameters of the GA during runtime using runtime adapter, when pre-determined iterations unattained. The probabilistic parameters are updated iteratively until an optimized schedule is attained.

    METHODS AND SYSTEMS FOR NEURAL ARCHITECTURE SEARCH
    9.
    发明公开
    METHODS AND SYSTEMS FOR NEURAL ARCHITECTURE SEARCH 审中-公开

    公开(公告)号：US20240354579A1

    公开(公告)日：2024-10-24

    申请号：US18732052

    申请日：2024-06-03

    Applicant: Swisscom AG

    Inventor： Yassine Benyahia ,  Kamil Bennani-Smires ,  Michael Baeriswyl ,  Claudiu Musat

    IPC: G06N3/086 ,  G06F18/24 ,  G06F40/20 ,  G06N3/045 ,  G06N3/048

    CPC classification number: G06N3/086 ,  G06F18/24 ,  G06F40/20 ,  G06N3/045 ,  G06N3/048

    Abstract: Methods and systems are provided for neural architecture search. In a system with suitable processing circuitry, a preferred model may be determined for performing a selected task, with the determining including obtaining a computational graph that includes a plurality of nodes and a corresponding plurality of weightings configured to scale input data into the nodes. The computational graph defines a first model and a second model with each of the models including a subgraph in the computational graph, with one or more of the plurality of weightings being shared between the first model and the second model. One or more weightings of each of the models may be updated based on training of each of the models to perform the selected task, and the preferred model may be identified based on an analysis of both models. A neural network for performing the selected task may be configured based on the preferred model.

Patent Agency Ranking