SISNR implementation

[krantiparida]

Hi, Thanks for the awesome work. I was going through the Si-SNR implementation but found that you have used negative of sdr here . If I have not grossly misunderstood anything, can you please let me know why you have used negative here? Thanks .

[haoheliu]

Hi Kranti,

Thanks for the interest of this repo. I add the negative sign in the Si-SNR because I once tried to use it as loss function. The negative sign turn loss minimization into Si-SNR maxmization. Hope this helps.

Haohe

[krantiparida]

Thanks for the prompt response. So, as I understand, if I need to evaluate using the implementation of Si-SNR then I have to remove the negative sign?

[haoheliu]

Exactly, you need to remove the negative sign if you use it for evaluation.

[krantiparida]

Thanks for the clarification.