17886381. TECHNIQUES FOR DOWNLOADING MODELS IN WIRELESS COMMUNICATIONS simplified abstract (QUALCOMM Incorporated)

TECHNIQUES FOR DOWNLOADING MODELS IN WIRELESS COMMUNICATIONS

Organization Name

QUALCOMM Incorporated

Inventor(s)

Rajeev Kumar of San Diego CA (US)

Gavin Bernard Horn of La Jolla CA (US)

Xipeng Zhu of San Diego CA (US)

Aziz Gholmieh of Del Mar CA (US)

TECHNIQUES FOR DOWNLOADING MODELS IN WIRELESS COMMUNICATIONS - A simplified explanation of the abstract

This abstract first appeared for US patent application 17886381 titled 'TECHNIQUES FOR DOWNLOADING MODELS IN WIRELESS COMMUNICATIONS

Simplified Explanation

The abstract describes a system for updating machine learning capabilities at a user equipment (UE) based on a list of supported models or model structures received from a network node, and downloading models from a model repository based on the updated capability.

• Receiving a list of supported models or model structures identifiers per machine learning function name or feature from a network node.
• Updating the capability at the UE based on the received list of supported models or model structures identifiers.
• Downloading one or more models from a model repository per machine learning function name or feature based on the updated capability and available resources at the UE.
• Transmitting the list of supported models or model structures identifiers and configuring the use of a model identifier for a specific machine learning function name or feature.

Potential Applications

• Enhancing machine learning capabilities on user equipment devices.
• Improving the efficiency of model selection and downloading in machine learning applications.

Problems Solved

• Updating machine learning capabilities based on network support.
• Efficiently downloading models based on device capabilities and resources.

Benefits

• Improved performance of machine learning applications on user equipment.
• Optimized use of available resources for model selection and downloading.

Original Abstract Submitted

Aspects described herein relate to receiving, from a network node, a list of supported models or model structures (MS) identifiers (IDs) per machine learning function name (MLFN) or machine learning feature (MLF) at the network node, updating a capability at the UE to an updated capability based on the list of supported models or MS IDs per MLFN or MLF at the network node, and downloading, at the UE and from a model repository, one or more models or MSs per MLFN or MLF based on the updated capability and available resources at the UE. Other aspects relate to transmitting the list of supported models or MS IDs and configuring use of a model or MS ID for a particular MLFN or MLF.