3 ms·
> TF [...] The final mean squared error is 0.0003. > Neural Designer [...] reaches a mean squared error of 0.023. > The following table summarizes the the[sic
by sva_ 4y ago
> TF [...] The final mean squared error is 0.0003.
> Neural Designer [...] reaches a mean squared error of 0.023.
> The following table summarizes the the[sic] most important metrics that the two machine learning platforms yielded .
[omits MSE]
They should train both to the same loss and then compare.
- BobbyJo 4y agoI'd like to know why there is a difference in the final loss at all. If the two networks had the same architecture, used the same loss function, and had random uniform initialization, then 1000 epochs should have them converging on very similar final loss values. Especially if one was able to converge to 3e-4.