Introduction - If you have any usage issues, please Google them yourself
These programs include the following aspects of a perceptron and ADALINE networks to identify the letters. Randomly selected initial weight vector, select the appropriate step length (ADALINE network), four inputs with the given training sample, training on the above two networks, respectively, until the network convergence 3. Selected Adaline Network different values , respectively, to draw the error curve, observe their variation . perceptron select the initial weight vector, calculated separately for each type of training samples to the hyperplane distance, to observe their similarities and differences 5. end of the training recognition ability of the test network (using the 100 samples tested, corresponds to each take 25 noisy deformation): 6. compare Adaline and single-neuron perceptron the classification results.