Hypergraph Query Generation from Discrete Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Analytics computing systems face challenges in generating database queries that accurately reflect discrete training data, as traditional machine learning models based on continuous data struggle to effectively utilize discrete training samples for querying database management systems.
Innovation Solution
Utilizing a knowledge base data structure represented as a hypergraph, the system generates subgraph queries based on positive and negative training samples to develop queries that accurately reflect desired data relationships, employing techniques like isomorphic subgraph identification, node unification, and hooking cycles to refine queries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional machine learning models based on continuous data are used, then the system can leverage established mathematical techniques, but the system cannot effectively utilize discrete training samples for generating accurate database queries
Solution Approach 1:
The patent transforms the discrete training data into a continuous representation space by encoding discrete concepts as nodes in a hypergraph and using continuous vector embeddings. This allows traditional continuous machine learning models to operate on discrete data by changing the parameter space from discrete categories to continuous vector representations, thereby resolving the incompatibility between discrete training data and continuous ML models.
Solution Approach 2:
The patent introduces a knowledge base hypergraph as an intermediary structure between discrete training samples and continuous machine learning models. Thehypergraph serves as a mediator that converts discrete conceptual relationships into a continuous graph representation, enabling seamless integration of discrete data with continuous learning algorithms while preserving the semantic relationships in the original discrete data.
2Quantity of substance
If the system generates queries to include all possible data relationships, then comprehensive data coverage is achieved, but unwanted or irrelevant data relationships are also included reducing query precision
Solution Approach 1:
The patent applies local quality by assigning different weights and importance levels to different edges and nodes in thehypergraph based on their relevance to the query objective. Instead of treating all data relationships uniformly, the system selectively emphasizes locally relevant relationships while suppressing irrelevant ones, thereby achieving precise query results without sacrificing comprehensive data coverage.
Solution Approach 2:
The patent implements feedback mechanisms where the system evaluates query results and adjusts the hypergraph structure or query parameters accordingly. By using feedback from query performance and validation against training samples, the system refines which data relationships to include or exclude, progressively improving query precision while maintaining adequate data coverage through iterative optimization.
Data Source
AI summary
Various examples are directed to systems and methods for generating a query of a database. An example method may comprise accessing a knowledge base data structure comprising plurality of nodes and a plurality of edges and accessing a plurality of training samples comprising a plurality of positive training samples and a plurality of negative training samples. The example method may also comprise determining a subgraph comprising a subset of the plurality of nodes and a subset of the plurality of edges, the subset of the plurality of nodes comprising at least two nodes that are also part of the plurality of training samples and executing a subgraph query of the knowledge base data structure, the subgraph query being based at least in part on the subgraph.