Quite possibly. The classification results were great, just wasn't good when trying to run things back through the network repeatedly. I did have issues that th

Quite possibly. The classification results were great, just wasn't good when trying to run things back through the network repeatedly. I did have issues that the learning rates reported in some of the original papers didn't match the ones in the released code.