100%(1)1 out of 1 people found this document helpful
This preview shows page 1 - 7 out of 153 pages.
Unsupervised Learning NetworksDr. K. GanesanDirector – TIFAC-CORE in Automotive Infotronics & Senior Professor in School of Information Technology and EngineeringVIT University, Vellore – 632 014[email protected]
Introduction•The second major learning paradigm is unsupervised learning. •Here there exists no feedback from the system (environment)to indicate the desired outputs of a network. •The network by itself should discover any relationships of interest, such as features, patterns, contours, correlations or categories, classificationin the input data and thereby translate the discovered relationships into outputs. •Such networks are also called self-organizing networks.
•Here the learning is based only upon the input data and is independent of the desired output data and also no error is calculated to train a network.•This type of learning is called unsupervised learning and the input data are called unlabeled data.•Here the net may respond to several output categories on training.•But only one of the several neurons has to respond. •The mechanism by which only one unit is chosen to respond is called competition.•The frequently used competition among group of neurons is called Winner-Takes-all.•Here only one neuron in the competing group will have a non-zero output signal when the competition is completed.
Clustering net•In a clustering net, there are as many input units as input vector components.•Since each output unit represents a cluster, the number of output units will limit the number of clustersthat can be formed.•The weight vector for an output unit in a clustering net is called exemplar or code-book vector for the input patterns, which the net has placed on that cluster.
Methods for determining the winner•There are 2 methods used for the determination of the winner unit.•Method 1•This method uses the squared Euclidean distance between the input vector and the weight vector, and chooses the unit whose weight vector has the smallest Euclidean distance from the input vector.•Method 2•This method uses the dot product of the input vector and the weight vector.•The dot product of an input vector with a given weight vector is the net input to the corresponding cluster unit.•The largest dot product corresponds to the smallest angle between the input and weight vectorsif they are both of unit length.
Example•A supervised learning can judge how similar a new input pattern is to typical patterns already seen, and the network gradually learns what similarity is. •The network may construct a set of axes along which to measure similarity to previous patterns, i.e it performs principal component analysis, clustering, adaptive vector quantization and feature mapping.