Artificial neural network

Winner Take All
1. Problem Definition
We have the words “AN” and “AM” which we need to recognize from a display of 16 segments with a medium noise or when a word is not correctly displayed

2.Proposed Solution
Using the JAVA technology we are going to do a GUI were is showed the neural network and in order to make the user feel like “the black box effect“ the user will enter the data andwith the implementation of the Winner Take All algorithm only the result is showed in the screen and the operations not.

3. Objectives
General Design and develop a program which has the ability ofrecognize the words “AN” and “AM” using a graphic interface for better user interaction and implementing the recognition algorithm named “Winner Take All” Specific Recognize the letters with 1 o 2segments missing The neuronal network need to have 3 layers The activation and inhibition values must be of 0.02 and -0.45 respectively The algorithm need to have a limit of 1000 iterations

4.Requirements
Functional Introduce the characteristics intuitively avoiding the use of plain text Separate the classes of the creation of the GUI showed to the user and the implementation of the algorithmNon-Functional In case that the user didn’t want to enter the data himself and only want to check the functionality of the program use test data or an example ready to work Modify the values likenumber of iterations or activation/inhibition weights

5. Flow Diagram

6. Use Case Diagram

Back Propagation
7. Problem Definition
We need a program that can learn to recognize the data inputsof the XOR table. X 0 0 1 1 y 0 1 0 1 XOR 0 1 1 0

8. Proposed Solution
Using the JAVA technology we are going to do a GUI were is showed the neural network and in order to make the user feel like“the black box effect“ the user will enter the data and with the implementation of the Back Propagation algorithm only the result is showed in the screen and the operations not.

9. Objectives...