ThreeLayerBP

A starter example for those new to Neural Networks.

Neural network is some very interesting stuff and has been gaining popularity through the last several years, it can be applied to many fields: Computer Vision, Data Mining, Automatic driving and so on. This is a most simple neural network used to approach a given linear or nonlinear function. Taking advantage of the architecture’s simplicity, you can use it as your first program implementing a simple neural network, either with keras, tensorflow or from scratch. Clone this repo and edit in the jupyter notebook.

git clone https://github.com/FengyangZhang/ThreeLayerBP.git
cd ThreeLayerBP
jupyter notebook

You may need to pip install numpy and matplotlib first, and keras / tensorflow if you like.

self implemented version

Open ThreeLayerBPfromScratch.ipynb and run the two cells to see the loss. You can change the implementations in neural_nets/NonLinearNet.py.

keras version

Open KerasThreeLayerBP.ipynb and run the cells, it’s pretty simple.

tensorflow version

Open TensorflowThreeLayerBP.ipynb, there’s also a Lenet implementation in TensorflowLenet.ipynb.