Radial Basis Function Networks is a type of artificial neural network that consists of three layers with feed-forward connections between the nodes, such as an input layer, a hidden layer with nonlinear radial basis activation function, and a linear output layer. RBFN is employed from the theory of functional approximation, which is utilized in many real-world applications. RBFN is a popular replacement for multi-layer perceptron. The significance of radial basis function networks is a universal approximation, better generalization, fast training, and quick learning speed. The main goal is to implement the input-output mapping using a linear combination of radially symmetric functions, and training an RBF network with linear outputs is accomplished in two stages. The first stage is unsupervised that accomplished by obtaining cluster centers of the training set input vectors by using k-means clustering, and the second stage consists of solving a set of linear equations, the solution of which can be obtained by a matrix inversion technique such as singular value decomposition or by least squares. The most application areas of radial basis function neural network are time series analysis, pattern modeling, image processing, speech recognition, adaptive equalization, classification, prediction, radar point-source location, fault detection, medical diagnosis, and many more. The future enhancements of the radial basis function network are a multi-layer RBF network with fuzzy information, the combination of RBF architecture with CNN for visual distribution of images, meta-heuristic optimization algorithm for RBFNN, automated cell nuclie identification using RBFNN for thyroid lesions cytology, and many more.