Neural Network Food Formula Generator
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for developing plant-based food alternatives struggle to accurately mimic the taste and texture of animal-based foods, relying on time-consuming and inefficient manual laboratory processes that often result in resource wastage.
Innovation Solution
A computer-implemented method using neural networks to generate plant-based food formulas by creating an ingredient vocabulary embedding matrix and a latent space with a probability distribution, allowing for the sampling and decoding of plant-based ingredients to mimic animal-based foods, ensuring only plant-based ingredients are used.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual laboratory work is used to develop food formulas by combining different ingredients and testing, then the process allows for direct sensory evaluation and adjustment, but it is time-consuming, inefficient, and wastes physical resources
Solution Approach 1:
The patent creates a digital twin or virtual model of the food formulation process using machine learning algorithms. Instead of physically combining and testing ingredients, the system uses a trained model to predict optimal ingredient combinations that match target sensory attributes, effectively copying the formulation process in silico before any physical production occurs.
Solution Approach 2:
The patent replaces manual laboratory work and physical ingredient testing with an automated machine learning system. The mechanical process of mixing, tasting, and adjusting formulas by hand is substituted with computational algorithms that analyze ingredient databases and predict optimal formulations based on learned patterns from training data.
2Reliability
If extensive manual testing of different ingredient combinations is performed, then the process can identify effective formulas, but it requires extensive time and physical resources
Solution Approach 1:
The patent performs preliminary action by training the machine learning model on extensive ingredient and recipe data before actual formulation occurs. The system pre-learns the relationships between ingredients and sensory outcomes, so when a new formulation is needed, the model can quickly predict optimal combinations without requiring extensive physical testing, thus reducing development time while maintaining reliability.
3Object-affected harmful factors
If plant-based ingredients are used to create alternatives to animal-based foods, then health and environmental benefits are achieved, but the taste and texture matching is insufficient
Solution Approach 1:
The patent uses parameter changes by adjusting ingredient ratios, processing conditions, and formulation parameters within the machine learning model to optimize sensory outcomes. The system can fine-tune multiple parameters simultaneously to achieve precise matching of taste, texture, and other sensory attributes, overcoming the limitations of simple plant-based substitutions.
Data Source
AI summary
Techniques to mimic a target food item using artificial intelligence are disclosed. A formula generator learns from open source and proprietary databases of ingredients and recipes. The formula generator is trained using features of the ingredients and using recipes. Given a target food item, the formula generator determines a formula that matches the given target food item and a score for the formula. The formula generator may generate numerous formulas that match the given target food item and may select an optimal formula from the generated formulas based on score.