Active Learning for Continual Learning and Sequential Decision-Making

Sequential decision-making is the key to active data collection for training data-driven surrogate models and design optimization for simulation-based design. By leveraging the quantified model uncertainty, I design algorithms to enable different types of active learning to identify the queries that can maximize the improvement of the system, thereby to maximize sampling efficiency in complex system design and Digital Twins. This stream of research contributes various engineering domains, including data fusion and design optimization using multi-fidelity data sources, integrating simulation and experimental data sources, but can also connect to autonomous discovery in material design. Future research directions including leveraging agentic systems to connect active learning algorithms with experimental/simulation platform for self-discovery.

Journal Articles

Wang, Z., Chen, Y.-P., Dolar, T., & Chen, W., “ARCO-BO: Adaptive Resource-aware COllaborative Bayesian Optimization for Heterogeneous Multi-Agent Design.” Journal of Mechanical Design, 148 (2026): 091703.

Chen, Y.-P., Wang, L., Comlek, Y., & Chen, W., “A Latent Variable Approach for Non-Hierarchical Multi-Fidelity Adaptive Sampling.” Computer Methods in Applied Mechanics and Engineering, 421 (2024), 116773.

Conference Talks

Chen, Y.-P., Wang, L., Comlek, Y., & Chen, W., “Data Fusion of Multi-fidelity Systems via Latent Variable Gaussian Process for Active Learning Applications.” 2nd ICMA MMLDE-CSET, Sep. 23, 2023, El Paso, Texas, USA

Chen, Y.-P., Wang, L., Comlek, Y., & Chen, W., “A Unified Adaptive Sampling Framework for Multi-Fidelity Modeling and Bayesian Optimization via Latent Variable Gaussian Process”, Society of Engineering (SES), Oct. 08, 2023, Minneapolis, Minnesota, USA