only noise generated

[jiqizaisikao]

I generated the sine wav with pitch 200HZ of 10 S wav file and trained wavenet with 300 steps ,and the loss down to 0.001,but when i geneated the wave file the out put seems like white noise! I tried it sevele times,all the output is noise,why? not enough trainning steps?but the loss is so small ,and the trianning data is only normal sine wav.

[devinroth]

Overtrained. It doesn’t work the way you think it does. You need lots of different data not just one sample. A minor deviation from the original sine way will generate noise. The neural network doesn’t know how to handle the deviation since you only trained it on one scenario.

Try training with a number of different sine waves of various frequencies and starts.

On Jan 20, 2018, at 8:21 PM, jiqizaisikao [email protected] wrote:

I generated the sin wav with pitch 200HZ of 10 S wav file and trained wavenet with 300 steps ,and the loss down to 0.001,but when i geneated the wave file the out put seems like white noise! I tried it sevele times,all the output is noise,why? not enough trainning steps?but the loss is so small ,end the trianning data is only normal sin wav.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ibab/tensorflow-wavenet/issues/322, or mute the thread https://github.com/notifications/unsubscribe-auth/AMaIEdAF2f1NQul6vkc0cGpUm-KBoGm6ks5tMrsugaJpZM4RlsRB.