Graph Embedding Model for Product Formulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI and machine learning systems in product development and design processes are limited by their reliance on general domains with few elements, ignoring ingredient amounts and additional features, and providing limited real-world suggestions and applications.
Innovation Solution
A computer-implemented method and system for representational learning of product formulas, which generates a directed graph from a set of product formulas, converts it into a weighted graph, and uses this graph to create an embedding model representing each ingredient with low-dimensional numerical vectors, enabling the generation of new product formulas and suggestions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If general domain AI systems are used for product development, then automation is achieved, but the system provides limited real-world suggestions and applications
Solution Approach 1:
The patent transforms product formulation data into graph embeddings, changing the parameter representation from traditional tabular formats to vector space representations. This enables the AI system to process complex ingredient relationships and generate meaningful product suggestions, resolving the contradiction between automation and real-world applicability.
Solution Approach 2:
The patent introduces graph embeddings as an intermediary layer between raw product formulation data and AI model processing. This intermediary transformation captures complex ingredient relationships and interactions, enabling the automated system to generate practically applicable product development suggestions.
2Device complexity
If traditional data representation methods are used, then simplicity is maintained, but ingredient amounts and additional features are ignored
Solution Approach 1:
The patent transitions from traditional flat data representation to multi-dimensional graph embeddings. This dimensional transformation allows the system to encode ingredient amounts, features, and relationships in a compressed vector space, preserving information while enabling efficient AI processing.
3Reliability
If product development relies on human expertise, then quality suggestions are provided, but the process takes years of human effort
Solution Approach 1:
The patent creates a digital twin of expert knowledge through graph embeddings learned from existing product formulations. This copied knowledge enables the AI system to generate high-quality product suggestions without requiring years of human expertise, significantly reducing development time while maintaining suggestion quality.
Data Source
AI summary
A method, system, and computer program product for representational learning of product formulas are provided. The method accesses a set of product formulas. Each product formula includes a set of ingredient tuples. A directed graph is generated from the set of product formulas. The directed graph including a node for each ingredient of the sets of ingredient tuples of the set of formulas. The method generates a weighted graph from the directed graph. The weighted graph has a weight assigned to each edge in the directed graph. The method generates an embedding model based on the directed graph. A set of embeddings is determined for the weighted graph where each node is represented with low-dimensional numerical vectors.