Inverse Learning: A Data-driven Framework to Infer Optimizations Models

摘要: We consider the problem of inferring optimal solutions and unknown parameters a partially-known constrained using set past decisions. assume that constraints original optimization are known while decisions objective to be inferred. In such situations, quality solution is evaluated in relation existing observations problem. A method previously used settings inverse optimization. This can infer utility functions decision-maker find based on these inferred indirectly. However, little effort has been made generalize methodology data-driven address solutions. this work, we present linear framework (Inverse Learning) aims an directly observed data validate our model dataset diet recommendation setting personalized diets for prediabetic patients with hypertension. Our results show obtains daily food intakes preserve trends providing range options providers. The proposed able both capture minimal perturbation from given and, at same time, achieve inherent objectives trade-off different metrics provide insights into how environments.