The partial order constraint between tasks is defined. Based upon these constraints, a group of originally independent tasks can be associated. The application of the partial order constraint is explored, a co-evolutionary multitask learning framework is presented, the learning process is a repeatedly process of independently evolution for each task and associated adjustment of taste. The application of the framework in the construction of the loss curves of pig chilling process is ansJyzed. Take into consideration of the partial order relation between low humid loss curve and mid humid loss curve, through a process of co-evolution, two reasonable curves can be constructed when there are only a few samples. The test result on four bench test functions suggests this approach is effective in more general condition.%定义任务之间的偏序限制,基于偏序限制可以联系原先独立的任务.分析偏序限制的应用,给出一个协同演化的多任务学习框架,它反复地通过各个任务的独立演化以寻优,通过联合调整以结合偏序限制.给出本框架在构建猪肉预冷损耗曲线过程中的应用:考虑了低湿损耗曲线与中湿损耗曲线间的偏序关系,利用协同演化,在样本.量很少时,也能获得合理的低温和申温损耗曲线.对于4个标准测试函数的测试显示了本策略对于一般问题的有效性.
展开▼
