Convolutional unsupervised learning?

[tempestatis]

Hello guys,
I want to train the filters in a convolutional network in a unsupervised manner without finetuning.
Afterwards the learned weights should be visualized.
I don't know how to define the solver.prototxt because I can't apply the autoencoder #330 example to my issue.
Thank you very much for your advice.

[shelhamer]

You have to go about defining your model and deciding how to train. The #330 should at least be useful as an example starting point.

If you have a precise question about how to define your model or solver configuration, please ask, but an explanation of convolutional unsupervised learning is for you to make for yourself.