English

Real-world Video Adaptation with Reinforcement Learning

Networking and Internet Architecture 2020-09-01 v1 Artificial Intelligence

Abstract

Client-side video players employ adaptive bitrate (ABR) algorithms to optimize user quality of experience (QoE). We evaluate recently proposed RL-based ABR methods in Facebook's web-based video streaming platform. Real-world ABR contains several challenges that requires customized designs beyond off-the-shelf RL algorithms -- we implement a scalable neural network architecture that supports videos with arbitrary bitrate encodings; we design a training method to cope with the variance resulting from the stochasticity in network conditions; and we leverage constrained Bayesian optimization for reward shaping in order to optimize the conflicting QoE objectives. In a week-long worldwide deployment with more than 30 million video streaming sessions, our RL approach outperforms the existing human-engineered ABR algorithms.

Keywords

Cite

@article{arxiv.2008.12858,
  title  = {Real-world Video Adaptation with Reinforcement Learning},
  author = {Hongzi Mao and Shannon Chen and Drew Dimmery and Shaun Singh and Drew Blaisdell and Yuandong Tian and Mohammad Alizadeh and Eytan Bakshy},
  journal= {arXiv preprint arXiv:2008.12858},
  year   = {2020}
}

Comments

Reinforcement Learning for Real Life (RL4RealLife) Workshop in the 36th International Conference on Machine Learning, Long Beach, California, USA, 2019

R2 v1 2026-06-23T18:10:31.101Z