lwestfall / NeuralNet

A simple implementation of a Neural Network modeled after Neilsen's book - but in .NET 7

NeuralNet

This is a basic proof of concept for a neural network implemented in .NET 7. It's heavily based on Michael Nielsen's excellent book Neural Networks and Deep Learning, with some minor twists to make it a little more object oriented.

The purpose of this project is simply to get more hands on experience with how neural networks work.

Setup

I'm assuming you have .NET 7 installed. If not, you can get it here.

1. Clone this repo
2. Download the MNIST dataset from here and extract it to the Datasets/digits folder

To Run

1. cd src/Sandbox
2. dotnet run

What's Next?

• Add a way to save and load trained networks
• Implement ReLU activation function
• Add convolutions
• Add unit testing
• Benchmarks and performance optimizations

References

http://neuralnetworksanddeeplearning.com/

https://github.com/tromgy/simple-neural-networks/blob/master/neural-network.ipynb

https://www.youtube.com/watch?v=aircAruvnKk&list=PLZHQObOWTQDNU6R1_67000Dx_ZCJB-3pi