about optimization process

[GaoitheFierce]

Hi,

I noticed that in the optimization process, you use self.X, but it seems to be replaced when a new batch is added. Does this mean that the optimization process effectively uses only one sample?

Here's the relevant code snippet for reference:

W = W_proximal_preprocess(W, self.X, self.dev)

Could you please confirm if my understanding is correct? If not, I'd appreciate it if you could provide more details on how self.X is handled during the optimization process.

Thank you!

[AozhongZhang]

Yes, you are right. We just use one sample for the proximal process. And we have tried 128 samples. The results are close to the result of one sample.