CSD 33903 by Ummu Rifqi https://padlet.com/ummurifqi09/CSD33903 Made with the strength to succeed en-us 2019-02-19 03:02:41 UTC 2026-01-28 16:45:16 UTC hello@padlet.com https://padlet-assets.s3.amazonaws.com/icons/Templeball.png REFERENCE ummurifqi09 https://padlet.com/ummurifqi09/CSD33903/wish/342008727 For preparation to test and final exam, kindly read : https://tinyurl.com/y5lnryco]]> 2019-03-16 14:22:56 UTC https://padlet.com/ummurifqi09/CSD33903/wish/342008727 Assignment 1 ummurifqi09 https://padlet.com/ummurifqi09/CSD33903/wish/342009170 is there anybody missed the submission?]]> 2019-03-16 14:27:32 UTC https://padlet.com/ummurifqi09/CSD33903/wish/342009170 Markah coursework ummurifqi09 https://padlet.com/ummurifqi09/CSD33903/wish/354678320 Markah akan dikemaskini dari masa ke semasa]]> 2019-04-27 14:47:02 UTC https://padlet.com/ummurifqi09/CSD33903/wish/354678320 TEST ANSWER SCHEMA ummurifqi09 https://padlet.com/ummurifqi09/CSD33903/wish/354678372 an question?

section A.. please ask those who came to the scheduled class on 24th april]]> 2019-04-27 14:47:54 UTC https://padlet.com/ummurifqi09/CSD33903/wish/354678372 Komen Assignment 1 https://padlet.com/ummurifqi09/CSD33903/wish/354738430 sila rujuk]]> 2019-04-28 05:52:52 UTC https://padlet.com/ummurifqi09/CSD33903/wish/354738430 https://padlet.com/ummurifqi09/CSD33903/wish/357063870 2019-05-06 06:47:51 UTC https://padlet.com/ummurifqi09/CSD33903/wish/357063870 Assignment 1 ummurifqi09 https://padlet.com/ummurifqi09/CSD33903/wish/483512786
Assignment 1 (CLO3 – 15%)

Training a dataset with basic neural network (supervised ANN)

  1. Choose a dataset (you can view the dataset in excel, .csv format, .dat format). Make sure your dataset has a class/ label. (you find it in UCI database, the number of dataset row should be in 100-200 only)
  2. Preprocess the dataset (either input or output) using normalization if needed.
  3. Draft the ANN consider with the dataset, identify input and output (class)
  4. Design your ANN layers and nodes in each layer ( how many input nodes, hidden nodes and output nodes, what is the bias value, threshold value, what is range of random weight for each nodes (and bias) the first cycle of training, what activation function do you use and its deactivation function for back propagation and learning rate. 
  5. Develop a program for neural network back propagation using JAVA/ JAVA script or PHP/Javascript.
  6. Train the program using different number hidden layer, different range of weight, different learning rate with the time training, how many epoch/cycle. Write in table. After the training process is over, generate several more points and show the shape of the function learned by plotting the error points.
  7. Test the result using test dataset. Write in table
What you are going to do in (3).

  1. Divide dataset in to training set and test dataset ( usually 70%- 30%)
  2. Import the training data set, consider the row/ column of attribute of the data
  3. Choose online or offline weight changes mode (consider this when you create loop for ANN cycle.
  4. Set stopping value of cycle (for example error >0.01)
  5. forward phase
    1. calculate all the hidden nodes
    2. activate all the hidden nodes using chosen activation function
    3. using activated value calculate output
    4. Find the error for each calculated (and activated) output. You maybe use different activation function that more simple than hidden layer.
  6. Back propagate phase
    1. Base on the error, find error gradient for output layer (by using deactivation function according the previous activation function used in forward propagation for each layer)
    2. Calculate weight changes (hidden layer to output) by calculating weight gradient of output layer
    3. Calculate error gradient for hidden layer.
    4. Calculate weight changes (input layer to hidden layer) by calculating weight gradient of hidden layer.
  7. Loop for forward and back propagation until value of error of 0.01
 

 

Submit report that include the following outline:

  1. Cover
  2. Project Group Member and Task
  3. Report Outline
  4. Introduction
  5. Design of ANN
-          The ANN architecture (of the designed ANN) based on item 4

  1. The result of training in table and explanation/justification
    1. Final weight
    2. How many epoch cycle
    3. How many times for training
    4. The test result  with test data
      1. Error
      2. Time
      3. accuracy
    5. Attachment of program
 

]]> 2020-03-31 03:58:03 UTC https://padlet.com/ummurifqi09/CSD33903/wish/483512786