Straight Forward Constructive Deep Learning Neural Network (SFC-DLNN) algorithm

Description

SFC-DLNN (Straight Forward Constructive Deep Learning Neural Network) is a supervised learning algorithm more specifically a Multi-Layer Perceptron. Their network is constructed dynamically, starting with a minimal topology (perceptron) then successively new units are trained and added to, creating a neural network with a more complex multi-layered topology. Each time a new unit (neuron) is added to the network under construction, the weights of the entire network are regenerated and finally a feature space is then created where the data is likely to be linearly separable. The algorithm advantages are quick learning; topology size delimitation and constructed structure conservation even if the training set changes. We obtained from SFC-DLNN learning system an accuracy and specificity of 83.5% from a simulated data set using the uniform law distribution.