Can different layers have different learning rates?

[Tgaaly]

Can different layers have different learning rates? This would be useful in the case of using pre-trained layers and fine-tuning new layers on top (ideally the training of the new layers should have a higher learning rate).

[Tgaaly]

After checking the code - different layers have to have the same learning rate. I did a pull request to allow this - Pull Request #143