Abstract: Methods and systems for training a neural combinatorial optimization (NCO) model having a processor and memory for performing a task having a target distribution. The NCO model is meta-trained to learn an efficient heuristic on a set of distributions. The meta-trained NCO model is then fine-tuned to specialize a learned heuristic for the target distribution.

Abstract: A historical task database relating vehicle rollout decisions, vehicle maintenance states and subsequent deteriorations is created. A pattern analyzer may use an item-set mining algorithm on the task database to recommend whether a vehicle with its current maintenance state should be deployed. A supervisor uses this recommendation to make a rollout decision. These decisions are added to the database. Heuristic rules are defined to determine if the rollout decision was correct. The system to learns when a supervisor continues to make costly rollout errors. The system also discovers combinations of defects that lead to a rapid deterioration and makes recommendations that the vehicle be sent for maintenance rather than being rolled out.

Abstract: A historical task database relating vehicle rollout decisions, vehicle maintenance states and subsequent deteriorations is created. A pattern analyzer may use an item-set mining algorithm on the task database to recommend whether a vehicle with its current maintenance state should be deployed. A supervisor uses this recommendation to make a rollout decision. These decisions are added to the database. Heuristic rules are defined to determine if the rollout decision was correct. The system to learns when a supervisor continues to make costly rollout errors. The system also discovers combinations of defects that lead to a rapid deterioration and makes recommendations that the vehicle be sent for maintenance rather than being rolled out.