EVOLVING FEED-FORWARD ARTIFICIAL NEURAL NETWORKS USING BINARY AND DENARY DATASET
The application of Artificial neural networks (ANNs) to various domain of research has been on the increase lately. The concept of ANN is born out of correlating natural neurons to artificial neurons in solving human problems. Neural networks are conceptualized to relate input to output. In this work, we developed a single-layer perceptron neural network model to classify a binary dataset and multilayered perceptron to classify denary dataset. Our Neural network models in this work, were trained using the binary and denary datasets repository of the Modified National Institute of Standards and Technology (MNIST). Our perceptron model which is also referred to as Linear Binary Classifier consist of four basic components namely Input values, Weights and Bias, Net sum and Activation Function. The model was able to classify the binary dataset with a near perfect linear decision boundary when bias was introduced. In the case of multi-layered perceptron, a non-linear decision boundary was obtained with the denary dataset. The results as presented showed that the model classification has a good accuracy in classifying the datasets.