Decision-Aware Training for Sample-Based Generative Models
2026-07-01 • Machine Learning
Machine Learning
AI summaryⓘ
The authors study how models that predict uncertain outcomes usually learn without considering how costly certain mistakes are in real-life decisions. They suggest a new way to train these models by adding a penalty that reflects the real cost of wrong predictions, alongside the usual method based on data fit. This approach helps the models focus more on avoiding costly errors while still giving a full range of possible predictions. They tested their method on both made-up and real datasets and found it improved decision-making where mistakes matter most.
probabilistic forecastingsample-based generative modelsenergy scoreproper scoring rulesdecision-aware trainingcost-sensitive learningdifferentiable decision lossprobabilistic modelsforecastingmodel training objectives
Authors
Kornelius Raeth, Nicole Ludwig
Abstract
Sample-based generative models are increasingly used for probabilistic forecasting in high-stakes decision settings, yet their training objectives are blind to the decision maker's cost structure. These models are commonly trained with strictly proper scoring rules, such as the energy score, which allocate their training signal in proportion to data density, with no awareness of where forecast errors are most costly for downstream decisions. We therefore propose decision-aware training for sample-based generative models, augmenting the energy score objective with a differentiable decision loss that directly penalises the cost incurred by acting on the model's forecast. This combined loss is theoretically grounded, as the decision loss is itself a proper scoring rule. We validate our method on one synthetic and two real-world tasks, showing targeted improvements in cost-sensitive regions while retaining full probabilistic forecasts.