Performance differences for similar bit rate

[Pascal Bacchus]

Hi  Johannes,

I trained a model for a given architecture and lambda parameter multiple time and the performances obtained are varying significantly for each model, up to 0.7 dB for similar bpp.

Is this a behaviour to be expected ?

I suppose it is because some models are stuck in local minima.

Thanks for your time.

Best,

Pascal

[Johannes Ballé]

Dear Pascal,

the results should not differ significantly. You may need to train your models for a longer time, or lower the learning rate to reduce the noise resulting from stochastic gradient descent.

Hope this helps!

Johannes

--
You received this message because you are subscribed to the Google Groups "tensorflow-compression" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tensorflow-compre...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/tensorflow-compression/28d11b3c-1b23-4363-bfd1-97af5b79940en%40googlegroups.com.