Don't really have a good grasp of how to put together a Pylearn2 model yet. Looking at the Interactive Pylearn2 notebook it should be fairly easy to put together a simple model in a notebook and slowly make it more complicated; running it at different stages to see what's happening.
Notebook would start with simple MLP model, then add convolutional layers, look at traces and decide when to add more layers or extensions; try to justify changing hyperparameters.
Not likely to result in our best model, but might at least be informative.
Don't really have a good grasp of how to put together a Pylearn2 model yet. Looking at the Interactive Pylearn2 notebook it should be fairly easy to put together a simple model in a notebook and slowly make it more complicated; running it at different stages to see what's happening.
Notebook would start with simple MLP model, then add convolutional layers, look at traces and decide when to add more layers or extensions; try to justify changing hyperparameters.
Not likely to result in our best model, but might at least be informative.