- Python 100%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
|
|
||
| .github/workflows | ||
| Active Leaning | ||
| cVAE | ||
| README.md | ||
| workflow.png | ||
FEs-Inverse-Design
This repository contains the source code and datasets for the research paper: "Active learning in latent spaces enables rapid inverse design of ferroelectric ceramics for energy storage".
🚀 Key Features
Conditional Variational Autoencoder (cVAE): Constructs a coupled search space that synergistically models chemical constraints and domain structure evolution.
-
Latent Space Optimization: Implements a two-stage multi-objective genetic algorithm (NSGA-II) to navigate the latent space for Pareto-optimal solutions.
-
Active Learning Surrogate: Uses symbolic regression and ensemble learning (CatBoost, XGBoost, etc.) to predict energy density and efficiency with uncertainty quantification.
-
Phase-Field Simulation: Generate reference domain structures for training.
📁Repository Structure
cVAE/: Code for training the Conditional Variational Autoencoder and reconstructing energy landscapes.
ActiveLearning/: Latent space optimization algorithm, Surrogate model as fitness function of optimization algorithm, uncertainty quantification.
README/: Docs
