Вы находитесь на странице: 1из 17

Neural networks lab manual

Prerequisite: 1. Normalised data from lecture session

1. Download Encog java. Go to https://code.google.com/p/encog-java/downloads/list and


download encog-workbench-3.2.0-release.zip

2. Unzip encog-workbench-3.2.0-release.zip

3. Double-click encog-workbench-3.2.0-executable.jar or use commandl line: java –jar encog-


workbench-3.2.0-executable.jar

4. Click new project and select a folder eg: desktop

5. Put the normalised data in Microsoft Excel


6. Open new Excel and paste the value

7. Save as.. CSV (comma delimited) at encog project folder ed:desktop


8. You should get a csv data on your encog folder

9. Verify the csv file by checking the contents it should shows number with a comma delimiter
10. Highlight the csv file and right click it
11. Select Export toTraining(EGB) and enter 5 in input and 1 at output count

12. Select OK and you should get next windows and leave all untick and selelct OK

13. You should get a new file xxxxx.egb


14. Double click to check the contents and should have 5 input and 1 ideal and 1 significant
15. Now select File-New File

16. Select Machine Learning Method and put the file name eg.xxxxxx and select OK

17. Select Feedforward Neural Network and OK and put 5 at Input and 1 at output
18. Click Add at Hidden layer count and put 1 and OK

19. Click small …. Button at Activation Function HIddeen and selelct ActicationSigmoid and OK
20. Do same with the Activation Function output like step 19. You will get this settings configuration
and OK

21. You will get new file xxxx.eg(BasicNetwork)


22. Check the file by doubleclick and you can get your neural network design

23. Select the Train button


24. You will get this windows and OK

25. Select BackPropagation (BPROP) and OK


26. Put Learning Rate 0.7 and Momentum 0.2 and OK

27. New tab will open


28. Click start and wait until finish train

29. Click close and save


30. Check new weight by click Weight

31. Now the new weights are trained


32. Verify the networks with highlight the network file and go to Tools Validation Chart
33. Click OK

34. You will get the graph


35. Click the Data tab and check the result should represent the nearest value to to the ideal data

Вам также может понравиться