SyMANTIC – Novel Symbolic Regression to Discover Accurate Models from Data
T2025-090
The Need
In many scientific and industrial fields, there is a critical need for interpretable and accurate models that can be derived from complex datasets. Traditional machine learning methods often produce black-box models that lack transparency and interpretability, making it difficult to understand and trust the results. There is a growing demand for tools that can generate simple, understandable, and accurate models to facilitate better decision-making and innovation.
The Technology
OSU researchers created SyMANTIC, an advanced symbolic regression method designed to discover interpretable and parsimonious mathematical models from data. It efficiently selects important features and balances accuracy with simplicity using a novel algorithm. SyMANTIC is capable of handling noisy and small datasets, making it versatile across various applications. The method leverages GPU acceleration for faster computations, ensuring timely results without compromising on model quality.
Commercial Applications
- Chemistry: Optimizing chemical manufacturing processes and quality control
- Financial Modeling: Creating interpretable models for risk assessment and investment strategies.
- Healthcare: Designing diagnostic tools and personalized treatment plans based on patient data.
- Environmental Monitoring: Modeling and predicting environmental changes and impacts.
Benefits/Advantages
- Interpretability: Produces simple and understandable models.
- Efficiency: Quickly identifies relevant features and generates models.
- Versatility: Effective with small and noisy datasets.
- Accuracy: Balances model simplicity with high predictive accuracy.
- Scalability: Utilizes GPU acceleration for faster processing.