- Add bias neuron
- Calculate error
- Backprop error
- Train - Neurons update their connect weight values (learning!)
- Simplify training method, use basic stochastic gradient descent
- Add training method or Trainer class
- Standardize Data set shape for training