| Author |
Message
|
| maximg |
 Posted: Fri Mar 28, 2008 5:14 am Post subject: Artificial Neural networks |
 |
|
Forum Freshman
Joined: 26 Sep 2006
Posts: 27
Location: Melbourne, Australia
|
I am hoping someone could help explain some questions I have about Artificial Neural networks.
Having recently been reading about ANN I have found in each article/tutorial they only describe the math’s side of things, which is fine for the most part but they tend to completely fail with making either a real world analogy of how such a system might be implemented or at least some more in-depth explanation that can be conceptualized.
For example:
I understand that in implementing a NN it would contain a series of “neurons” that would have inputs and outputs; along with associated weights that would affect the output such as shown below.
| Code: |
Weight Weight
|| ||
INPUT -----------------------------------------------------OUTPUT
|
What I fail to understand is
• How those weights are to be applied to the input to create an output.
• How you can actually use such a system and apply it to a real world application.
o How do I break my problem up (whatever it might be) so that it can be input into the NN?
If anyone could guide me in the right direction about these questions I would be very grateful.
_________________
"THE ULTIMATE MEASURE OF A MAN IS NOT WHERE HE STANDS IN MOMENTS OF COMFORT
AND CONVENIENCE, BUT WHERE HE STANDS AT TIMES OF CHALLENGE AND CONTROVERSY."
Dr. Martin Luther King Jr |
|
| Back to top |
|
 |
| sunshinewarrior |
| Posted: Fri Mar 28, 2008 9:10 am Post subject: |
|
|
Forum Ph.D.
Joined: 26 Sep 2007
Posts: 974
Location: London
|
Best explanation I saw of neural nets and weighting was by Stephen Pinker in How the Mind Works. I'm not au fait enough with it, however, to attempt an independent explanation.
Here are some things I remember.
1. There are usually multiple connections and allowance for multiple inputs. I do not have the facility to represent it schematically, alas.
2. The weights all contribute through the multiple connections, thereby allowing a specific output - for instance if one of the inputs was 4 and it went through a 'neuron' that gave 4 a weight of 0 then it could output through a 'neuron' that represented "less than 5".
Dunno if this helps, but there you are... |
|
| Back to top |
|
|
| nikomaster |
| Posted: Sat Mar 29, 2008 10:17 am Post subject: |
|
|
Forum Freshman
Joined: 09 Nov 2007
Posts: 4
|
Weights are like constant that helps to modulate the input, in biological systems the weight is represented by the synapsis where depending of the size of the synapsis reponses the output of the neuron,in the same way , the mathematical model using a number that multiplies the input, if you are actually a programmer, maybe you are not able to understand how this can be used in the real world, but I assure you it has a wide range of applications, to understand Neural Networks you must understand how the biological neural systems works, remember that is a sum of all inputs where each input is multiplied by a weight and the output is decided by an activation function, is like life it self, a living being is determinated by the sum of it's acts where each act is motivated with a factor and the output is the result what that individue is, is not that hard to understand, think about a lot people working together in a task each person have a set of skills and the sum of the action of all the people in the team has a result= the output, and this output is determinated by te capacities of each person= the weight and the spectations of the output=the activation functions.
Try to learn more about Neural Networks |
|
| Back to top |
|
|
|
|
|
