Hi Lingkai, I have some questions when I use this to do regression task.
There are three arguements I don't know how to set:
The "target_scale", "self.delta" and "self.sigma", how could I set it for my own dataset.
About SDE-Net training and test detection:
When "not training_diffusion" in SDENet, we can get the output of mean and sigma, but what's the meaning of the "mean" and "sigma", Is this the mean or sigma value of target? But if so, the target value is a single value, how could it have mean and sigma value?
How could I get aleatoric uncertainty:
I've noticed that in "test_detection_sde.py", it will write out a file with model uncertainty. So if I want to write out inputs uncertainty, I just have to write the total Mean value. Is this right?
Actually, when I'm training my dataset, some metrics are not good, I just use the default setting of network, I want to know whether there is some tuning I can do to improve my network's performance.
Hi Lingkai, I have some questions when I use this to do regression task.
Actually, when I'm training my dataset, some metrics are not good, I just use the default setting of network, I want to know whether there is some tuning I can do to improve my network's performance.