The output neurons of the network compete with each other and only one neuron wins at the same time.
Ii. Rules of Learning
The learning rule of competitive neural network is a kohonen learning rule which is developed by the inner star rule.
4.SOM Learning Algorithms
- Set Variable: x=[x1,x2,x3,..., XM] is an input sample, each sample is an m-dimensional vector. Ωi (k) =[ωi1 (k), Ωi2 (k),..., Ωin (k)] is a weight vector between the I input node and the output neuron
- Initialization: Weights are initialized with smaller random values, and normalization of input vectors and weights is done
X ' = x/| | x| |
Ω ' I (k) =ωi (k)/| | Ωi (k) | |
|| x| | and | | Ωi (k) | | Euclidean norm of input vector and weight vector, respectively
- Sample Input Network: sample and weight vector do the inner product, the output neuron with the largest inner product value wins the competition and is credited as the winning neuron
- Update weights: Neurons in the neighborhood of the winning neuron topology are updated with the rules of the star
Ω (k+1) =ω (k) +η (x-ω (k))
- Update the learning rate η and topological neighborhood, and re-normalized the weight value after learning
- Determine whether convergence. If the center changes very little or reaches a predetermined number of iterations, the end algorithm
Self-organizing competitive neural network-matlab