rbf_networks_classification

Simple RBF Neural Networks for Two-class Classification Example

It is a simple example of using RBF neural network to deal with two-class classification tasks. This task is implemented by tensorflow. The directory contains four python files:

• kmeans.py, which is implemented to find the centre vectors of hidden neurons.
• rbf.py is the core model to achieve the RBF neural network.
• execute.py, the main function to load data in ./data/ directory, kmeans.py and rbf.py, then train and test.
• validation_mlp.py, since the testing dataset in ./data/ directory does not contain label information, so this multi-layer perceptron method is used to train a model and generate the testing results, then compare with the derived results via defined RBF neural network.

For the explanation and details of RBF networks, referring the following articles:

• RBF Network
• RBF network for backpropagation supervised training

To train and test the RBF model, run:

$ python execute.py

To train and test the multi-layer perceptron model, run:

$ python validation_mlp.py