Character recognition using HAM (Neural Network)

Neural Network using Auto Associative memory method to store 5 characters

A Hopfield Network has the following architecture:
◮ Recurrent network, weights Wij
◮ Symmetric weights, i.e. Wij= Wji
◮ All neurons can act as input units and all units are output units
◮ It’s a dynamical system (more precisely “attractor network”):
◮ It’s possible to store memory items in the weights W of the network and use it as associative memory
Pros:
◮ Very simple model
◮ Nice mathematical analysis possible (also for capacity)
Cons:
◮ Dynamics of the system are constrained to fixed points
◮ No storage of time series
◮ Low capacity
Reference:
http://www.igi.tugraz.at/lehre/NNB/SS10/Lecture_Hopfield_nets.pdf
Related Examples:
1. Car detection from images
https://in.mathworks.com/matlabcentral/fileexchange/63161-adaboost--pca--capstone-project-

2. Perceptron Learning (Neural Networks)
https://in.mathworks.com/matlabcentral/fileexchange/63046-perceptron-learning

3. Hebbian Learning (Neural Networks)
https://in.mathworks.com/matlabcentral/fileexchange/63045-hebbian-learning

4. Delta Learning rule, Widrow-Hoff Learning rule (Artificial Neural Networks)
https://in.mathworks.com/matlabcentral/fileexchange/63050-delta-learning--widrow-hoff-learning