My Neural Network

This neural network can also be called Multilayer Perceptron. The activation function is sigmoid.

How to use this repo

Command line instruction: we recommend using python3 to run this program The program includes two .py files: preProcessing.py and NeNet.py.

Pre-Process

the input parameters to the preProcessing.py file are:

• complete input path of the raw dataset, or instead we stored the url of the three following dataset, you can put the name for that particular dataset:
  • ds1: https://archive.ics.uci.edu/ml/machine-learning-databases/car/car.data
  • ds2: https://archive.ics.uci.edu/ml/machine-learning-databases/iris/iris.data
  • ds3: https://archive.ics.uci.edu/ml/machine-learning-databases/adult/adult.data
• complete output path of the pre-processed dataset. For example 'postProcessed.csv' can be a path for the output file.

for example

python3 preProcessing.py ds1 'postProcessed.csv'

The above would imply that the training dataset is 'ds1' which is the first dataset listed above. The output path is 'currentDirectory/postProcessed.csv'

Back-propagation

The input parameters to the NeNet.py are as follows:

• input dataset – a complete path the post-processed input dataset which you specfied for the output path of the preProcessing.py

• training percent – percentage of the dataset to be used for training

• maximum_iterations – Maximum number of iterations that your algorithm will run. This parameter is used so that your program terminates in a reasonable time.

• number of hidden layers

• number of neurons in each hidden layer

for example

python3 NeNet.py 'postProcessed.csv' 80 4000 2 10 10

The above would imply that the dataset is 'postProcessed.csv', the percent of the dataset to be used for training is 80%, the maximum number of iterations is 4000, and there are 2 hidden layers with (10, 10) neurons. Your program would have to initialize the weights randomly