Neural Network Feature (Wish) List

# Neural Network Feature (Wish) List
## These are the basics:
- [x] Basic 2-layer network with bias
- [x] Activation functions which reuse the activation on the backwards pass (sigmoid, tanh, ReLU)
- [x] MSE cost function
- [x] Adjustable learning rate
- [ ] Additional examples and tests
See #41 and #66 and #76.

## These would be interesting to add:
- [ ] Multiple hidden layers
Per #61.
- [ ] Semi-arbitrary activation functions
Per #70 and #75.
- [ ] Arbitrary cost functions
See [here](https://stats.stackexchange.com/questions/154879/a-list-of-cost-functions-used-in-neural-networks-alongside-applications).
- [ ] Automatically adapting learning rate (Momentum)
Per #65. Also see [here](https://distill.pub/2017/momentum/) via [here](https://ml4a.github.io/ml4a/how_neural_networks_are_trained/).
- [ ] Multiple initial weighting strategies
See [here](https://prateekvjoshi.com/2016/03/29/understanding-xavier-initialization-in-deep-neural-networks/).
- [ ] Convolution layers
See [here](https://towardsdatascience.com/building-a-deep-neural-net-in-google-sheets-49cdaf466da0). Also see [here](http://cs231n.github.io/convolutional-networks/) via [here](http://cs231n.github.io/).
- [ ] Simple RNN
See [here](https://gist.github.com/karpathy/d4dee566867f8291f086) via [here](http://karpathy.github.io/2015/05/21/rnn-effectiveness/).
- [ ] Advanced optimization
See [here](https://towardsdatascience.com/neural-networks-with-optimal-accuracy-and-speed-in-their-predictions-d2cdc3b21b50).

This issue is a broad wish list and names no files, tests, or entry points. Start by selecting one unchecked item and reading the referenced issues and linked resources. Define that item's scope and completion criteria before attempting work.

Written by the indexing model from the issue text.