How to code up Neural Networks ?

I am new to Neural Networks. I have to code up whatever is written in below image.

Is there a need of nprtool to do that?

[Explorer]

Can I do it using nprtool? If yes, how?

Answer 1

[Greg Heath]

I suggest using a standard classifier with patternnet, one hidden layer, and target columns from eye(23).

Use as many defaults as possible.

Search using

 greg patternnet

for examples.

Hope this helps.

Thank you for formally accepting my answer

Greg

[Explorer]

I have descriptor vector of 1x360 size and for feature matching I would like to do same as written in image of research paper.

Please tell me stepwise how to do that in MATLAB.

I don't have much knowledge of Neural Networks. Its not my subject.

Answer 2

[Greg Heath]

I am confused by your question

 "Is there a need of nprtool to do that?

Please clarify that statement.

Comment

The figure shows a 3-layer feed-forward multilayer perceptron neural network with three neuron layer outputs ai (i=1:3).

The input box indicates a layer of non-neuron fan-in-units.

There are two hidden neuron layers and an output neuron layer.

One hidden neuron layer is sufficient for a universal approximator.

One hidden layer FFMLPs are the "standard" configuration

The standard output for a 23-class classifier/pattern-recognizer is a 23 node output layer. The net is then trained with column vectors from the 23-dimensional unit matrix eye(23).

So, are you claiming that the extra hidden layer and single output learns faster and better than the standard classifier with one hidden layer and 23 outputs?

[Explorer]

I am confused by your question

 "Is there a need of nprtool to do that?

Please clarify that statement. _________________________________________________________________________

nprtool = neural pattern recognition tool

Will this tool be required to do whatever is said in above image. _________________________________________________________________________

I am not claiming. I am taking follow a research paper for my sign language recognition project. _________________________________________________________________________

The above images is of a research paper, F. E. Luis-Pérez, F. Trujillo-Romero, W. Martinez-Velazco, Control of a service robot using the Mexican Sign Language, 10th Mexican International Conference on Artificial Intelligence (MICAI 2011), Advances in Soft Computing, Lecture Notes in Artificial Intelligence, Springer, Vol. 7095, Proceedings, Part II, pp. 419-430, Puebla, Mexico, Nov. 26-Dec. 4, 2011, ISBN: 978-3-642-25329-4. ISSN: 0302-9743.

Answer 3

[Greg Heath]

Once you have the input matrix of N I-dimensional column vectors and target matrix of N c-dimensional column unit vectors you can use the code from one of my patternnet postings.

Searchwith

 greg patternnet

Hope this helps.

Thank you for formally accepting my answer

Greg