Data Efficient Surrogate Modeling for Engineering Design: Ensemble-free Batch Mode Deep Active Learning for Regression

Overview

Engineering design optimization requires expensive simulations (CFD, FEA). This work develops an ensemble-free batch mode deep active learning method for building accurate surrogate models with minimal training data — critical for making AI-driven design practical.

Key Contributions

• Ensemble-free approach — avoids the computational overhead of maintaining model ensembles
• Batch mode — selects multiple informative samples per iteration for parallel simulation
• Regression-focused — designed for continuous engineering metrics, not just classification
• Demonstrated on UUV hull design and other engineering benchmarks