Visualization of Artificial Neural Networks
Neural network visualization methods
Selected visualization methods have to be incorporated into a human-machine interface (HMI) providing visual information of the learning process of a Multilayer Perceptron type neural network (MLP). Proposed HMI has to enabled a user to comprehend the learning process of MLPs based on the visual information and thus incorporate his ideas and knowledge into the learning process.
- Overview of visualization methods is coming soon!
Reduction of visual information
Visualization of moderate sized MLPs presents an overwhelming amount of visual information to the user. Our method to reduce the user fatigue is based on clustering of Response Function Plots of hidden neurons. Then the user is presented only with the representatives of clusters. A scale of clustering algorithms is available to perform the task of clustering. The performance of Kohonen network, Growing Neural Gas (GNG), GNG with Utility factor (GNG-U) and Hybrid GNG (GNG-H) was tested.
Appropriate parameters for each algorithm were selected based on evaluation of following experiments. One iteration of each algorithm means presentation of randomly chosen sequence of the training patterns. There are two sets of training patterns representing responses of neurons in first hidden layers of a trained MLP.