README improvements

Author: gbuesing

README.md

@@ -1,11 +1,21 @@
 Neural Net (in Ruby)
 ===
 
-A [feedforward neural network](http://en.wikipedia.org/wiki/Feedforward_neural_network) with [resilient backpropagation](http://en.wikipedia.org/wiki/Rprop) (Rprop). It's ~250 loc, with no external dependencies.
+A [feedforward neural network](http://en.wikipedia.org/wiki/Feedforward_neural_network) with [resilient backpropagation](http://en.wikipedia.org/wiki/Rprop) (Rprop). It's ~250 loc, 100% Ruby, with no external dependencies.
 
-Run ```iris.rb``` to see an example of solving a simple classification problem: predict the species of iris flower based on sepal and petal size.
+This implementation trains significantly faster than [ai4r](https://github.com/SergioFierens/ai4r)'s backpropagation neural network, mainly because the Rprop training algorithm implemented here is much faster than the non-batch backpropagation algorithm used in ai4r.
 
-Sources and inspirations:
+However, this implementation is significantly slower than [ruby-fann](https://github.com/tangledpath/ruby-fann), which wraps the FANN library, written in C. If you're looking for something production-ready, use ruby-fann.
+
+
+Examples
+---
+- ```iris.rb```: solves a simple classification problem: predict the species of iris flower based on sepal and petal size.
+- ```mnist.rb```: performs OCR on handwritten digits. Requires download of MNIST dataset; see instructions at top of file.
+
+
+Sources and inspirations
+---
 
 - [Introduction to the Math of Neural Networks](http://www.amazon.com/Introduction-Math-Neural-Networks-Heaton-ebook/dp/B00845UQL6)
 - [Thoughtful Machine Learning: A Test-Driven Approach](http://www.amazon.com/Thoughtful-Machine-Learning-Test-Driven-Approach/dp/1449374069)