add learning rate function to allow decaying learning rate

[skaae]

Allows the user to specify the learning rate as a function of the epoch number.

Adds test that show how to use exponentially decaying learning rate

I'll add the same functionality for mementum if you accept the pull request.

btw is there a reason the ReLU pull request are not integrated?

[rasmusbergpalm]

Please properly remove all references to nn.scaning_learningRate.
Your new stuff depreceates it.

[rasmusbergpalm]

Awesome!
Only one small comment then I'll merge.
I'll look at the ReLU PR now

[rasmusbergpalm]

If you're referring to the #85 and #86 then it's because there are unresolved issues with those PRs, see the comments. If you'd resolve the comments then I'd be happy to merge them.

[gzbin365]

您好，邮件我已经收到了，我会尽快处理。

[wangruinan]

Thank you. Your email is received and will be handled as soon as possible！

[JustDoSomethong]

你好，我是黄俊嘉，邮件已到，我会尽快查收。