20230505-MTOA: Multitasking agents improve their average and maximum accuracy when tasks overlap.

This archive contains the results of a multi-agent simulation experiment [1] carried out with Lazy lavender [2] environment.Experiment Label: 20230505-MTOAExperiment design: Agents specialize by accepting or not to play a task.Experiment setting: Agents are trained with respect to different tasks and then coordinate upon acting on them. Each time they disagree, one agent adapts its knowledge with respect to the current task.Hypotheses: Agents will improve their accuracy more on tasks they choose to play.Detailed information can be found in index.html or notebook.ipynb.[1] https://sake.re/20230505-MTOA[2] https://gitlab.inria.fr/moex/lazylav/本存档包含基于Lazy lavender[2]环境开展的多智能体仿真（multi-agent simulation）实验[1]的结果。 实验标签：20230505-MTOA 实验设计：智能体（Agent）通过接受或拒绝任务实现专业化分工。 实验设置：智能体先针对各类任务完成训练，随后在执行任务时开展协作协调；每当出现意见分歧时，其中一名智能体将基于当前任务调整自身的知识模型。 实验假设：智能体在其主动选择承接的任务上，准确率提升幅度将更为显著。 详细信息可查阅index.html或notebook.ipynb。 [1] https://sake.re/20230505-MTOA [2] https://gitlab.inria.fr/moex/lazylav/