A Review of Supervised and Unsupervised Learning Techniques of Artificial Neural Network from Statistical Perspective

Artificial Neural Networks are richly connected networks of simple elements modeled on biological processes. Neural net models are specified by the net topology, node characteristics and training or learning rules. Learning in ANN requires the existence of a systematic procedure for adjusting the network weights to produce the output. There are three major learning paradigms, each corresponding to a particular abstract learning task. These are supervised learning, unsupervised learning and reinforcement learning. In supervised learning or associative learning, the network is trained by providing it with an input and matching output patterns whereas in unsupervised learning or selforganization an output unit is trained to respond to clusters of pattern within the input. It is generally assumed that neural net classifiers are non-parametric and make weaker assumptions concerning the shapes of the underlying distributions than traditional classifiers. In fact, neural networks involve exactly the same sort of distributional assumptions as statistical models. There are many research papers based on the comparative study of neural networks and statistics. A literature review of the supervised and unsupervised techniques of ANN from statistical perspective is presented in this paper. Also an attempt has been made to compare some supervised and unsupervised learning algorithms of ANN with statistical classification methods. It is also proposed to study the statistical features in the artificial neural network algorithms which are usually ignored by neural network researchers.