基于元学习的自适应视频流算法  

Adaptive video streaming algorithm based on meta-learning

作　　者：易令 李泽平[1] YI Ling;LI Ze-ping(School of Computer Science and Technology,Guizhou University,Guiyang 550025,China)

机构地区：[1]贵州大学计算机科学与技术学院,贵州贵阳550025

出　　处：《计算机工程与设计》2023年第3期641-647,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61462014)。

摘　　要：针对现有的码率自适应(adaptive bitrate,ABR)算法存在控制规则简单,不能有效提升用户体验质量(quality of experience,QoE),提出一种基于元学习的LABR(reinforcement learning based ABR)算法。采用策略梯度训练策略网络,利用元学习(meta-learning)方法学习基线(baseline)函数来减少因网络吞吐量差异产生的方差,进一步提高模型的准确性和鲁棒性;通过在策略函数中加入熵损失方法提高累计期望奖励值。实验结果表明,LABR算法具有泛化性与鲁棒性,能有效提高用户的视频体验质量。Aiming at the existing adaptive bitrate(ABR)algorithm,which has simple control rules and cannot effectively improve the quality of experience(QoE),a LABR(reinforcement learning based ABR)based on meta-learning was proposed.The policy gradient method was used to train the policy network,and the meta-learning method was used to learn the baseline function to reduce the variance caused by the difference in network throughput,and the accuracy and robustness of the model were further improved.By adding the entropy loss method to the policy function,the cumulative expected reward value was increased.Expe-rimental results show that the LABR algorithm has generalization and robustness,and can effectively improve the user’s video experience quality.

关 键 词：码率自适应算法 体验质量 元学习 策略梯度 基线 熵损失 期望奖励 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]