Analogical Reasoning System Using Neural Network Multiplexing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing analogical reasoning systems face limitations in efficiently handling complex inferences due to oversimplification of semantics, reliance on syntax, and inefficient computation times, making them unsuitable for real-world applications.
Innovation Solution
A high-performance general-purpose analogical reasoning system utilizing a neural network with symbolic substitution, object/predicate units, proposition units, sub-proposition units, and semantic units to find correspondences and infer solutions through non-temporal activation multiplexing and controlled activation flow, enabling efficient and versatile analogical reasoning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional analogical reasoning systems (SME, ACT-R, VAE) are used, then semantic information is oversimplified or syntax is relied upon, but the system fails to handle complex inferences effectively
Solution Approach 1:
The patent combines multiple types of units (object units, predicate units, proposition units, sub-proposition units, and semantic units) into a composite neural network structure. This composite architecture integrates both syntactic and semantic processing capabilities, allowing the system to handle complex inferences effectively while maintaining versatility in semantic information processing.
2Productivity
If existing reasoning systems are used, then computation time is excessive, but real-world applications require efficient processing
Solution Approach 1:
The patent segments the reasoning process into distinct unit types (object units, predicate units, proposition units, sub-proposition units, and semantic units) that can be processed independently and in parallel. This segmentation allows for more efficient computation by distributing the processing load across multiple specialized units, reducing overall computation time while maintaining high productivity.
3Ease of operation
If path-mapping algorithm of ACT-R is used, then only one path is considered at a time, but this introduces errors and fails to consider joint roles of objects
Solution Approach 1:
The patent creates a universal neural network structure where proposition units and sub-proposition units can represent multiple paths and joint roles simultaneously. Unlike ACT-R's single-path approach, this multi-functional architecture allows the system to consider multiple inference paths and the joint roles of objects across different paths, improving inference accuracy while maintaining algorithmic simplicity through the unified neural network framework.
Data Source
AI summary
The present invention relates to a general-purpose analogical reasoning system. More specifically, the present invention relates to a high-performance, semantic-based hybrid architecture for analogical reasoning, capable of finding correspondences between a novel situation and a known situation using relational symmetries, object similarities, or a combination of the two. The system is a high-performance symbolic connectionist model which multiplexes activation across a non-temporal dimension and uses controlled activation flow based on an analogical network structure. The system uses incremental inference to stop inference early for object correspondence, uses initial mappings to constrain future mappings, uses inferred mappings to synchronize activation, and independent mapping based on roles, superficial similarity, or composites.