.png){kind=icon}
Return
Dynamic multi-objective causal Bayesian optimization: Adaptive learning for complex systems under interventional constraints
.png){kind=icon}
PRE
AI
DOI：10.1016/j.knosys.2025.114971.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Real-world optimization problems increasingly involve complex systems where temporal dynamics, causal relationships, and competing objectives interact simultaneously. Traditional Bayesian Optimization (BO) approaches handle these challenges separately, limiting their effectiveness in dynamic environments with interventional constraints. Here, we present Dynamic Multi-objective Causal Bayesian Optimization (DMC-BO), a unified framework that integrates causal discovery with adaptive multi-objective learning to enable efficient optimization under temporal and cost constraints. The framework employs a dynamic acquisition function that leverages causal graphs to identify minimal intervention sets while adapting to system changes over time. Through causal-aware dominance criteria and cost-sensitive exploration strategies, DMC-BO maintains Pareto-optimal solutions while respecting budgetary limitations. Experimental validation across synthetic and real-world datasets (healthcare optimization and economic policy coordination) demonstrates that our approach achieves 35–56 % fewer interventions and 38–41 % lower costs compared to static baselines while maintaining robust performance under partial observability.
Journal
K
IF:
7.6
Papers:
1.3W
Citations:
4.5W
Organization
No organization information available
Cited Papers
.png){kind=icon}
ShareSave
ShareSave
ShareSave
View more