NOMADxzy/RL_ABR_SR

强化学习，决策码率、超分辨率

RL_ABR_SR

Re-implementation of existing neural ABR algorithms with Pytorch, MAML.

User guide

To Train

python main.py

To Test

python client.py

Note that the original version of Pensieve using asynchronous advantage actor-critic algorithm (A3C) to train the policy, which can only implementated on CPU. Our A2C version removes the asynchronous setting and use GPU to accelerate the speed of NNs training.

Further improvements are ongoing...