Decision Focused Learning

Decision-focused learning (DFL) integrates machine learning (ML) prediction and optimization to improve decision-making under uncertainty, directly optimizing for the quality of decisions rather than solely prediction accuracy. Current research focuses on developing efficient algorithms and model architectures, such as graph neural networks and bilevel optimization, to address challenges in various application domains including portfolio optimization, marketing, and public health intervention planning. This approach offers significant advantages over traditional "predict-then-optimize" methods by aligning the learning objective with the ultimate decision-making goal, leading to improved performance and robustness in real-world applications. The field is actively exploring robust loss functions and scalable methods to broaden the applicability of DFL to increasingly complex problems.