search

hongzimao / pensieve

Neural Adaptive Video Streaming with Pensieve (SIGCOMM '17)
http://web.mit.edu/pensieve/
MIT License
525 stars 281 forks source link

Range of bandwidth the linear model is trained on? #8

Closed Zahaib closed 7 years ago

Zahaib commented 7 years ago

Hey,

I noticed that the bandwidth traces you guys provided. The average BW across those traces is mostly less than 3000Kbps. Is this the same range of bandwidth the linear QoE model is trained for or has it also been trained for higher bandwidths like between 7000Kbps-9000Kbps?

I ask this because I may need to evaluate Pensieve on slightly higher bandwidths and would like to make sure if I need to retrain for it or not.

Thanks.

hongzimao commented 7 years ago

The traces we used is mostly less than 4.5Mbps. The reason (originally argued by MPC paper, which we follow) is to create challenging environment for bitrate decisions (ABR can be straightforward for too large or too small bandwidth).

Yes, I think you need to retrain Pensieve in your environment for fair comparison.

Zahaib commented 7 years ago

Makes sense, thanks.

yun-nam commented 7 years ago

Hi Hongzi, Thanks for your great work. I am wondering whether the traces you provided is subset of your traces or full set of traces you used in the paper for training and testing..

Thanks

Yun

hongzimao commented 7 years ago

Should be full. We also provided the links to raw traces and the way we processed them in traces/.