Concise Vector Representation for Edge User Profiles

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional service systems face storage and processing overheads due to large volumes of computed attributes from user event sequence data, leading to inefficient use of resources and reduced capabilities in machine learning-based services at edge systems.

Innovation Solution

A user representation model generates a concise vector representation of user event sequence data, reducing storage and computational requirements by applying a task-specific, multitask, or task-agnostic learning model, which is trained using a task prediction model to replace multiple computed attributes, allowing for efficient storage and processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If user event sequence data is stored in full detail at hub and edge systems, then complete user profile information is available, but storage overhead and processing resources are excessively consumed

Engineering Contradiction:

Improveuser profile information completenessVSAvoidstorage overhead

Core Design Contradiction:

Loss of informationVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential features from user event sequence data by generating fixed-size vector representations (embeddings) that capture user behavior patterns. Instead of storing complete event sequences, the system extracts condensed vector representations that preserve the necessary information for machine learning services while dramatically reducing storage requirements at both hub and edge systems.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms user event sequence data from its original high-dimensional format into fixed-size vector representations through neural network encoding. This parameter transformation converts variable-length event sequences into standardized vectors with predetermined sizes, enabling efficient storage and processing while maintaining the essential information needed for user profiling and machine learning services.

Inventive Principle:

Principle #35Parameter changes

2Reliability

If computed attributes from user event sequence data are stored extensively, then accurate machine learning services can be provided, but processing efficiency and resource utilization deteriorate

Engineering Contradiction:

Improvemachine learning service accuracyVSAvoidprocessing efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent performs preliminary processing of user event sequence data by generating fixed-size vector representations at the hub system before data is distributed to edge systems. This advance preparation of condensed user profiles enables edge systems to perform machine learning services efficiently without needing to process the original voluminous event sequences, thereby improving processing efficiency while maintaining service accuracy.

Inventive Principle:

Principle #10Preliminary action

3Adaptability or versatility

If large volumes of user event data are transmitted and stored at edge systems, then local machine learning services can operate, but resource constraints at edge systems are exceeded

Engineering Contradiction:

Improveedge system service capabilityVSAvoiddata storage capacity

Core Design Contradiction:

Adaptability or versatilityVSVolume of stationary object

Solution Approach 1:

The patent creates condensed copies of user event sequence data in the form of fixed-size vector representations. Instead of copying and storing the original voluminous event data at edge systems, the system generates compact vector copies that preserve the essential user profile information. These vector copies enable edge systems to provide machine learning services with minimal storage requirements, overcoming resource constraints while maintaining service versatility.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS12182829B2Generating concise and common user representations for edge systems from event sequence data stored on hub systems

Publication Date: 2024.12.31 ADOBE INC

AI summary

A system includes a representation generator subsystem configured to execute a user representation model and a task prediction model to generate a user representation for a user. The user representation model receives user event sequence data comprises a sequence of user interactions with the system. The task prediction model is configured to train the user representation model. The user representation includes a vector of a predetermined size that represents the user event sequence data and is generated by applying the trained user representation model to the user event sequence data. A storage requirement of the user representation is less than a storage space requirement of the user event sequence data. The system includes a data store configured for storing the user representation in a user profile associated with the user.