Exploiting Multi-Task Learning to Achieve Effective Transfer Deep Reinforcement Learning in Elastic Optical Networks

Exploiting Multi-Task Learning to Achieve Effective Transfer Deep Reinforcement Learning in Elastic Optical Networks
复制标题

DOI：
10.1364/ofc.2020.m1b.3
复制
发表时间：
2020-03
期刊：
2020 Optical Fiber Communications Conference and Exhibition (OFC)
影响因子：
--
通讯作者：
Xiaoliang Chen;R. Proietti;Che-Yu Liu;Zuqing Zhu;S. Yoo
Xiaoliang Chen;R. Proietti;Che-Yu Liu;Zuqing Zhu;S. Yoo
中科院分区：
其他
文献类型：
--
作者：
Xiaoliang Chen;R. Proietti;Che-Yu Liu;Zuqing Zhu;S. Yoo

文献摘要

被引文献

相似文献

我们提出了一种多任务学习辅助知识转移方法，用于EONs中有效和可扩展的深度强化学习。RMSA的案例研究表明，该方法可以将学习时间减少4倍，阻塞概率降低17.7%。
We propose a multi-task-learning-aided knowledge transferring approach for effective and scalable deep reinforcement learning in EONs. Case studies with RMSA show that this approach can achieve ∼ 4× learning time reduction and ∼ 17.7% lower blocking probability.