Sparse Kernel Model for Material Discovery with Limited Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods face challenges in predicting chemical formulas with desired properties from a chemical database, especially when insufficient training data is available, and struggle with converting feature vectors into new materials efficiently.
Innovation Solution
A machine learning framework employing a sparse kernel model is used to predict property values from feature vectors, selecting existing materials with high predicted property values and basis materials with large reaction magnitudes to generate new material candidates as variants, leveraging deep learning techniques for efficient material discovery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional methods are used to predict chemical formulas with desired properties, then the process can be performed with standard techniques, but prediction accuracy deteriorates when training data is insufficient
Solution Approach 1:
The patent introduces a sparse kernel model as an intermediary between the feature vectors and property predictions. This model acts as a mediator that can generalize from limited training data by leveraging the sparse kernel representation, thereby maintaining prediction accuracy even when training data is insufficient
Solution Approach 2:
The patent transforms the chemical formulas into feature vectors and uses a sparse kernel model with adjustable parameters to predict properties. By changing the parameterization approach from direct conventional prediction to sparse kernel-based prediction, the system achieves better accuracy with limited data
2Productivity
If conventional methods convert feature vectors into new materials, then the process follows traditional approaches, but computational efficiency deteriorates
Solution Approach 1:
The patent replaces conventional mechanical/iterative conversion methods with a machine learning-based sparse kernel model. This substitution enables direct prediction of property values from feature vectors without requiring complex iterative conversion processes, thereby improving computational efficiency
Solution Approach 2:
The patent performs preliminary extraction of feature vectors from chemical formulas before prediction. By pre-processing the chemical data into structured feature vectors, the system prepares the input in advance for efficient sparse kernel model prediction, reducing overall computational complexity
Data Source
AI summary
A computer-implemented method is presented for discovering new material candidates from a chemical database. The method includes extracting a feature vector from a chemical formula, learning a prediction model for predicting property values from the feature vector with a sparse kernel model employing the chemical database, selecting an existing material from a list of existing materials sorted in descending order based on the predicted property values by the prediction model learned in the learning step, selecting a basis material from a list of basis materials sorted in descending order of absolute reaction magnitudes to the selected existing material, and generating the new material candidates as variants of the selected existing material with consideration of the selected basis material.