A20 – Neural Networks

This activity is similar to the last two activities in which we try to identify what class an object belongs to. This time we are going to use neural networks which compose of three parts. These are the input layer, hidden layer and the output layer. The input layer contains the characteristics in which the objects can be distinguished and the output layer would tell which class the object belongs to. I would use the same classes I used in activity 19. The two classes are the quail eggs and the squid balls. Neural networks needs to train themselves with the objects and their classification in order to be more efficient in classifying the objects later on. The code was given by Jeric. The modified code is given below for my data set. The code is shown below.

--------------------------------------------------------------------------------------------------

// Simple NN that learns 'and' logic

// ensure the same starting point each time
rand('seed',0);

// network def.
// - neurons per layer, including input
//3 neurons in the input layer, 8 in the hidden layer and 1 in the ouput layer
N = [3,8,1];

// inputs
x = train = fscanfMat("train.txt")';// Training Set

// targets, 0 if squidballs, 1 if quail eggs
t = [0 0 0 0 1 1 1 1 ];
// learning rate is 4 and 0 is the threshold for the error tolerated by the network
lp = [4,0];

W = ann_FF_init(N);

// 1000 training cyles
T = 1000;
W = ann_FF_Std_online(x,t,N,W,lp,T);
//x is the training t is the output W is the initialized weights,
//N is the NN architecture, lp is the learning rate and T is the number of iterations

// full run
j=ann_FF_run(x,N,W) //the network N was tested using x as the test set, and W as the weights of the connections

--------------------------------------------------------------------------------------------------

The results were
[5.34E-03 7.41E-02 1.44E-02 8.34E-03 0.99132471 0.952957436 0.97008475 0.981131133]

Rounding off shows a 100% accuracy. The weakness of this method is it needs to train first a lot of times and also needs a number of samples in order to distinguish between classes. Our brain is far more advance since we can automatically tell which is which even if we just saw it once.

I give myself a grade of 10 in this activity since I was able to understand and employ neural networks in classifying objects. Thank you for Jeric for the code.