Parts 4 + 5

[emilygrabowski]

Part 4: ANN example, keras

1. It would be useful to include discussion of more complex architectures of neural nets (such as RNNs and CNNs, which may be useful to folks looking to apply this to specific issues
2. Dataset: same as above (handwriting), but with higher resolution. Ideally we would use the exact same data for both cases
3. In the “Prepare data” section, include some additional justification and description of the methods, especially since I expect that most folks will be accustomed to the sklearn style of preprocessing.
4. Similarly for the “Build ANN” section, include specific information about how models are built in keras, some of the common layers, and how to choose parameters
5. Ideally, I think the model we build in keras should parallel the one in sklearn, in order to highlight the differences in coding between the two packages.

---

Part 5: Load model/ visualize

1. Combine this with the previous notebook, since it deals with the same model