Deep-Learning Entity Vector Clustering for Social Network Data Access

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current deep-learning models for social networking systems fail to effectively differentiate between entities a user has interacted with and those they haven't, leading to inefficient data storage and retrieval, particularly in geographically distributed databases, and lack methods to optimize data access times and reduce data traffic.

Innovation Solution

A computer-implemented method using a deep-learning model to generate vector representations of entities, allowing for the calculation of similarity scores and updating of entity representations based on user interactions, which identifies relevant entities by clustering them in a multi-dimensional embedding space and storing them closer to the user for faster access.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Volume of stationary object

If data is stored in geographically distributed data centers without considering user relevance, then data storage capacity is improved, but data access time increases and data traffic increases

Engineering Contradiction:
Improvedata storage capacityVSAvoiddata access time
Core Design Contradiction:
Volume of stationary objectVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-computing vector representations of entities and pre-identifying relevant entities for users before actual data access occurs. The system calculates similarity scores and determines which entities to store closer to users in advance, so that when users access data, the relevant information is already positioned for fast retrieval, eliminating the need for real-time computation during access operations.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements local quality by creating different storage locations for different types of data based on user relevance. Specifically, it identifies relevant entities and stores them in local data centers closer to users, while non-relevant entities remain in remote data centers. This spatial differentiation of data quality and location based on local user needs reduces access time for relevant data while maintaining overall storage capacity.

Inventive Principle:
Principle #3Local quality

2Productivity

If all entities are stored in remote data centers, then data storage efficiency is improved, but data traffic increases when serving user requests

Engineering Contradiction:
Improvedata storage efficiencyVSAvoiddata traffic
Core Design Contradiction:
ProductivityVSLoss of energy

Solution Approach 1:

The patent applies local quality by differentiating storage locations based on data relevance to users. Relevant entities are stored in local data centers closer to users, while non-relevant entities remain in remote data centers. This creates a heterogeneous storage architecture where data quality and location vary according to local user needs, reducing the amount of data traffic required to serve user requests while maintaining storage efficiency.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent implements segmentation by dividing the unified data storage system into multiple segments: local data centers and remote data centers. Each segment serves a specific function - local centers store user-relevant data for fast access, while remote centers store less frequently accessed data. This segmentation of the storage system reduces overall data traffic by keeping frequently accessed data locally, while maintaining efficient use of total storage capacity across all segments.

Inventive Principle:
Principle #1Segmentation

3Device complexity

If deep-learning models classify all entities uniformly, then model simplicity is maintained, but the ability to differentiate between user-interacted and non-interacted entities is lost

Engineering Contradiction:
Improvemodel complexityVSAvoidentity differentiation accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent applies dimensionality change by transforming entity data into vector representations in a multi-dimensional embedding space. Instead of using simple categorical classification, the system represents entities as vectors with multiple dimensions capturing various attributes and relationships. This dimensional transformation enables the deep-learning model to differentiate between user-interacted and non-interacted entities by computing similarity scores in this rich vector space, significantly improving differentiation accuracy while maintaining reasonable model complexity.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP3188086B1Identifying entities using a deep-learning model
Publication Date: 2020.02.19 META PLATFORMS INC
  • EP3188086B1 patent drawingFigure 1
  • EP3188086B1 patent drawingFigure 2
  • EP3188086B1 patent drawingFigure 3

AI summary

In one embodiment, a method includes accessing a first set of entities, with which a user has interacted, and a second set of entities in a social-networking system. A first set of vector representations of the first set of entities are determined using a deep-learning model. A target entity is selected from the first set of entities, and the vector representation of the target entity is removed from the first set. The remaining vector representations in the first set are combined to determine a vector representation of the user. A second set of vector representations of the second set of entities are determined using the deep-learning model. Similarity scores are computed between the user and each of the target entity and the entities in the second set of entities. Vector representations of entities in the second set of entities are updated based on the similarity scores using the deep-learning model.