17545573. FEDERATED MACHINE LEARNING BASED ON PARTIALLY SECURED SPATIO-TEMPORAL DATA simplified abstract (INTERNATIONAL BUSINESS MACHINES CORPORATION)

From WikiPatents
Jump to navigation Jump to search

FEDERATED MACHINE LEARNING BASED ON PARTIALLY SECURED SPATIO-TEMPORAL DATA

Organization Name

INTERNATIONAL BUSINESS MACHINES CORPORATION

Inventor(s)

Lokesh Nagalapatti of Chennai (IN)

Sambaran Bandyopadhyay of Hooghly (IN)

Ruhi Sharma Mittal of Bengaluru (IN)

Ramasuri Narayanam of ANDHRA PRADESH (IN)

FEDERATED MACHINE LEARNING BASED ON PARTIALLY SECURED SPATIO-TEMPORAL DATA - A simplified explanation of the abstract

This abstract first appeared for US patent application 17545573 titled 'FEDERATED MACHINE LEARNING BASED ON PARTIALLY SECURED SPATIO-TEMPORAL DATA

Simplified Explanation

The abstract describes a method, system, and computer program for federated machine learning using partially secured spatio-temporal data. The method involves obtaining temporal data from multiple client devices, where some data is encoded private data and some is public data. A spatio-temporal graph is generated, representing the client devices, by identifying similar nodes based on the public data and adding edges between them. Encoders of the client devices are aligned based on the spatio-temporal graph.

  • The method involves federated machine learning using partially secured spatio-temporal data.
  • Temporal data is obtained from multiple client devices, including encoded private data and public data.
  • A spatio-temporal graph is generated, representing the client devices.
  • Similar nodes are identified based on the public data, and edges are added between them in the graph.
  • Encoders of the client devices are aligned based on the spatio-temporal graph.

Potential Applications

  • This technology can be applied in scenarios where multiple client devices generate spatio-temporal data, such as IoT devices, mobile devices, or sensor networks.
  • It can be used in applications that require federated machine learning, where data privacy and security are important considerations.
  • The method can be applied in various domains, including healthcare, transportation, smart cities, and environmental monitoring.

Problems Solved

  • The method addresses the challenge of federated machine learning with partially secured spatio-temporal data, where some data is private and needs to be protected.
  • It solves the problem of aligning encoders of distributed client devices to ensure consistent and accurate machine learning models.
  • The method also solves the problem of identifying similar nodes in the spatio-temporal graph based on public data, allowing for efficient data analysis and collaboration.

Benefits

  • The method allows for federated machine learning while preserving data privacy and security by encoding private data.
  • It enables collaboration and knowledge sharing among distributed client devices by generating a spatio-temporal graph and aligning encoders.
  • The method improves the accuracy and consistency of machine learning models by aligning encoders based on the identified similar nodes in the graph.


Original Abstract Submitted

Methods, systems, and computer program products for federated machine learning based on partially secured spatio-temporal data are provided herein. A computer-implemented method includes obtaining temporal data from a plurality of distributed client devices in conjunction with a federated machine learning process, wherein at least a portion of the data comprises encoded private data and at least a portion of the data is public data; generating a spatio-temporal graph comprising nodes representing the plurality of distributed client devices, wherein the generating comprises identifying at least one pair of similar nodes based at least in part on the public data and adding an edge to the spatio-temporal graph between the pair of similar nodes; and aligning encoders of at least two of the distributed client devices based on the spatio-temporal graph.

Retrieved from "http://wikipatents.org/index.php?title=17545573._FEDERATED_MACHINE_LEARNING_BASED_ON_PARTIALLY_SECURED_SPATIO-TEMPORAL_DATA_simplified_abstract_(INTERNATIONAL_BUSINESS_MACHINES_CORPORATION)&oldid=25754"