Scientific Machine Learning / Materials Science
Thin-Film Scientific ML
PyTorch and Gaussian-process models for predicting thin-film thickness from molecular properties.
- Institution
- Garcia Center for Polymers at Engineering Interfaces
- Role
- Researcher
- Timeline
- Jun — Dec 2024
- Status
- Research complete
Computational materials research using four-dimensional manifold learning and Gaussian Process Regression.
PyTorchGaussian Process RegressionManifold learning
01Polymer data024D representation03Gaussian process04Thickness prediction
Thin-film behavior depends on nonlinear relationships between molecular properties and material thickness.
Use four-dimensional manifold learning in PyTorch and Gaussian Process Regression to model molecular weight and film thickness.
Dhruva implemented the representation-learning and predictive-modeling workflow.
The model predicted commercial polystyrene semiconductor-wafer thickness.
Scientific ML is strongest when representation choices and uncertainty remain interpretable.