Good-Enough Model Spaces
We include a preliminary selection of code demonstrating GEMS in practice. More complete code (corresponding to all three datasets) is forthcoming.
GEMS ( Neural Network).ipynb: A Jupyter notebook containing a single run (end-to-end) for GEMS with a two layer neural network on MNIST (5 nodes)
GEMS (Convex).ipynb: A Jupyter notebook containing a single run (end-to-end) for GEMS with a logistic model on MNIST (5 nodes)
data/: partitioned MNIST data used for the notebook tasks
utils/: utility files for training models/executing GEMS