HI... I am new to work with neural networks in matlab...

i need to train the neural network for accumulator output like

[001; 010; 011; 100; 101; 001; 010; 011; 100; 101; 001; 010; 011; 100; 101; 001; 010; 011; 100; 101; 001; 010; 011; 100; 101; 001; 010; 011; 100; 101; 001; 010; 011; 100; 101; 110; 110; 110; 110; 110; 110; 110; 001; 001; 010; 010; 011; 011; 100; 100; 101; 101; 110; 110; 111; 111]

this is a 57*1 matrix

h0w the target file must be?

i need to get output 1 for this sequence of inputs...

and for the rest of the input values i need to get 0??

and how to specify the target file??

[Greg Heath]

If you are trying to isolate one 57 dimensional vector in a sea of 8^57-1 alternatives why not just use a distance measure?

The problem does not appear to be suitable for anything else.

Hope this helps.

Greg

[prabu shankar]

@greg heath: sir actually i have input matrix of size 57*3..in neural network clustering what is the format to make the inputs available.... it is saying no inputs selected ..tats what i meant this time.... hope ur reply soon!

[Greg Heath]

You did not clearly define the problem: 1. Binary? 2. Classification or clustering? 3. Number of alternatives (N = 8^57-1 or less?) 4. Which MATLAB function? 5. Representative codeand error message?

[prabu shankar]

@Greg Heath: sir, i trained the network with 56*3 matrix..

[1  1  1
10  10  10
11  11  11
100  100  100
101  101  101
1  1  1
10  10  10
11  11  11
100  100  100
101  101  101
1  1  1
10  10  10
11  11  11
100  100  100
101  101  101
1  1  1
10  10  10
11  11  11
100  100  100
101  101  101
1  1  1
10  10  10
11  11  11
100  100  100
101  101  101
1  1  1
10  10  10
11  11  11
100  100  100
101  101  101
1  1  1
10  10  10
11  11  11
100  100  100
101  101  101
110  110  110
110  110  110
110  110  110
110  110  110
110  110  110
110  110  110
110  110  110
1  1  1
1  1  1
10  10  10
10  10  10
11  11  11
11  11  11
100  100  100
100  100  100
101  101  101
101  101  101
110  110  110
110  110  110
111  111  111
111  111  111 ]...

since i need to get value 1 for this pattern..i had made the target file to be 56*1 matrix..all the elements have the value 1.

1.yes sir,i do have binary values..

2.since my main aim is for pattern reecognition.. i hope to use pattern recognition tool..

3.alternatives are not as much sir ..if a pattern is given other than the trained pattern i need to get a value 0..how to test the netork with othr patterns sir??

Answer 1

[Greg Heath]

You want to single out one 56 dimensional binary vector out of 8^56-1 alternatives.

Calculate the hamming distance between the the target vector and the input. If the distance is 0, the input is the target, and the output is 1. Otherwise it is not and the output is zero.

Nothing magical.

Not necessary to use any NN MATLAB functions.

Hope this is clear (since it is essentially a repeat of prior advice).

Greg