Meta-learning helps a model adapt to new tasks quickly. 元学习帮助模型快速适应新任务。
By using meta-learning, the system leverages experience from many related tasks to learn a good initialization, so it can fine-tune effectively with only a few examples. 通过元学习,系统从许多相关任务的经验中学习到一个良好的初始化,因此只需少量样本也能有效微调。