Classnotes

Artificial Intelligence

eXercise #2

Artificial Neural Networks (ANNS)

1. COUNTER NETWORK: The first 8 counting numbers are represented in binary as follows:
   
   

   Binary	Decimal		Binary	Decimal
   0000	0		0101	5
   0001	1		0110	6
   0010	2		0111	7
   0011	3		1000	8
   0100	4

   Create a neural network (call it COUNT.NET on your H: drive) that will accept the binary representation and correctly output the decimal equivalent. Train your neural network and demonstrate it's effectiveness to your teacher.
   
   
   
2. The TC network : a simple Optical Character Recognition Neural Network

   Consider the following optical characters as all of the regular orientations of the characters 't' and 'c'.

   1 1 1 0 1 0 0 1 0

   0 0 1 1 1 1 0 0 1

   0 1 0 0 1 0 1 1 1

   1 0 0 1 1 1 1 0 0

   1 1 1 1 0 0 1 1 1

   1 1 1 1 0 1 1 0 1

   1 1 1 0 0 1 1 1 1

   1 0 1 1 0 1 1 1 1

   We could specify a rule for each of the orientations of each letter as follows:
   
   

   IF top left cell is on 
      and top centre is on 
      and top right is on 
      and middle left is off 
      and middle centre is on 
      and middle right is off 
      and bottom left is off 
      and bottom centre is on 
      and bottom right is off 
   THEN the letter is a 'T'.
   

   INPUTS (1 = On, 0 = Off)									OUTPUT
   Top Left	Top Middle	Top Right	Middle Left	Middle Middle	Middle Right	Bottom Left	Bottom Middle	Bottom Right	Letter
   1	1	1	0	1	0	0	1	0	10 = T
   1	1	1	1	0	0	1	1	1	20 = C

   (note NSHELL can only currently deal with NUMERIC DATA)

   This system has 9 INPUTS, 1 OUTPUT, 8 RULES (the total number of rows in the matrix).

   CREATE a Neural Network according the specifications above. SAVE the network under the name VISION.NET in your home directory.
   TRAIN the network - this may take a number of minutes (aim to get the minimum error (it should be possible to get 0 error).
   RUN the network using some of the KNOWN data, to see if it classifies them correctly.
   TEST the network on some unknown data like the following:
   

   1	1	1
   1	0	1
   1	1	1

   ..this should be more like a 'c' than a 't'
   
   MAINTAIN the network by training it to recognise T, C and L
   Add to its training with letters H, I and X
   
   

3. Design and create a Neural Network to deal with XOR (the truth table follows)
   
   

   Input 1	Input 2	Output
   true	true	false
   true	false	true
   false	true	true
   false	false	false

   
   NOTE: the system has 2 inputs, 1 output and 4 rules. Allow the system plenty of time to train, then test it correctly classifies test data.