Timeline for Where has RSquare from tensorflow_addons.metrics gone and what's the appropriate replacement in model.compile(…) of keras?
Current License: CC BY-SA 4.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 18 at 12:06 | comment | added | BsAxUbx5KoQDEpCAqSffwGy554PSah | Your code gives me an erroron tensorflow 2.13: "ValueError: Tensor conversion requested dtype int32 for Tensor with dtype float32: <tf.Tensor: shape=(), dtype=float32, numpy=0.0>". As reported by another user on Stack Overflow. | |
| Jul 27 at 9:53 | comment | added | Oxbowerce |
I just used matplotlib to plot the predictions against the labels. Why not use mean squared error as your loss function? R-squared is a reformulation of the MSE, so minimizing the MSE is equal to maximizing the R-squared.
|
|
| Jul 27 at 8:59 | vote | accept | AlMa1r | ||
| Jul 27 at 8:52 | comment | added | AlMa1r | As for the MeanSquaredError, I'm unsure what to use instead. Ideally, I'd use 1-R2, but I don't find a built-it loss function for it. | |
| Jul 27 at 8:51 | comment | added | AlMa1r | Thx! How have you plot it ? | |
| Jul 27 at 8:40 | vote | accept | AlMa1r | ||
| Jul 27 at 8:54 | |||||
| Jul 27 at 8:36 | vote | accept | AlMa1r | ||
| Jul 27 at 8:38 | |||||
| Jul 26 at 18:33 | history | answered | Oxbowerce | CC BY-SA 4.0 |