w2 w Input layer Hidden layers Output layer Figure 7.A typical three-layer feed-forward network architecture. on error-correction,Hebbian,and competitive learning. tures.However,each learning algorithm is designed for Learning rules based on error-correction can be used for training a specific architecture.Therefore,when we dis training feed-forward networks,while Hebbian learning cuss a learning algorithm,a particular network archi- rules have been used for all types of network architec- tecture association is implied.Each algorithm can Table 2.Well-known learning algorithms. Paradigm Learning rule Architecture Learning algorithm Task Supervised Error-correction Single-or Perceptron Pattern classification multilayer learning algorithms Function approximation perceptron Back-propagation Prediction,control Adaline and Madaline Boltzmann Recurrent Boltzmann learning Pattern classification aigorithm Hebbian Multilayer feed- Linear discriminant Data analysis forward analysis Pattern classification Competitive Competitive Learning vector Within-class quantization categorization Data compression ART network ARTMap Pattern classification Within-class categorization Unsupervised Error-correction Multilayer feed- Sammon's projection Data analysis forward Hebbian Feed-forward or Principal component Data analysis competitive analysis Data compression Hopfield Network Associative memory Associative memory learning Competitive Competitive Vector quantization Categorization Data compression Kohonen's SOM Kohonen's SOM Categorization Data analysis ART networks ART1.ART2 Categorization Hybrid Error-correction RBF network RBF learning Pattern classification and competitive algorithm Function approximation Prediction,control 38 ComputerFigure 7. A typical three-layer feed-forward network architecture. on error-correction, Hebbian, and competitive learning. Learning rules based on error-correction can be used for training feed-forward networks, while Hebbian learning rules have been used for all types of network architec￾tures. However, each learning algorithm is designed for training a specific architecture. Therefore, when we dis￾cuss a learning algorithm, a particular network archi￾tecture association is implied. Each algorithm can Computer