The Jupyter-Notebook-File Your_first_neural_network.ipynb includes my solution to the Predicting-Bike-Sharing-Patterns-Project in the Deeplearning-Nanodegree.
The task is to use a neural network to predict the number of bike sharing users per day.
The idea is to attack this task by training a fully connected neural network feeding it historical data of bike-sharing user behaviour. The neural network is built by hand completely. This means that the forward- and back-propagation is implemented directly without the usage of any deeplearning library.
The network has one fully connected hidden layer and one fully connected output layer.
- Number of hidden nodes: 20
- Number of output nodes: 1
- Learning rate: 0.6
- Number of training iterations: 4000